Re: Money and mouse clicks
On Thu, Sep 19, 2002 at 10:15:19AM +0100, Brad Bollenbach wrote: > > So my question is, if you can get a decent enough website going to make > thousands and thousands of people click on a "Send" button (Hotmail) or > a "Google Search" button, how then do you have any hope of actually > making money off that? Errr. I wouldn't use those two as examples. Google's model has *never* been to make money off the Internet. They use the Internet as the test-bed for their technologies which they sell at corporate rates to corporate customers (eg to index and search company intranets). The Google Answers service is the first to actually make money directly off their Internet visitors - and it works by taking a %age of what is effectively a peer-peer service. Hotmail is, purely and simply a loss-leader, and I suspect it will never be anything else. I'm not sure that clicks = money is a sufficient way to think about it, but that's JMO. Ben
Re: networks and scalability
[Warning: Long and possibly overly techie at the end] On Tue, Sep 24, 2002 at 12:51:22AM +0100, Simon Wistow wrote: > > I've been working on this http://thegestalt.org/simon/remove this bit > between these two slashes/gestalt/ > > The basic idea is to try and map the topology of the London > nu-media-and-related-industries community. > > Anyway, one of the (non enforced) rules is that you don't link from > yourself to another person unless you play the 'bodily fluid' get out > rule - if you're blood relatives or seeing somebody. Instead you link > through other entities such as houses, workplaces, universities and > online cabals (such as, err, this mailing list). The reasoning behind > this is that I thought that it would be more scaleable - if I had a link > to all my friends and all my friends had a link to all my other friends > then it could get very messy, very quickly. > > However a friend of mine argued that having everyone you knew listed > explicitly would be better since you met everybody you know from a > situation and so its only going to create extra links by linking through > the situation as a third party. > > Does that make sense (in a purely semantic sense)? You already know that I think the hierarchical system is a better approach. Not least because the linkages which are possible have some kind of semantic meaning. Eg, it is meaningless to say that a place attends another place, or an online cabal dated a university. This provides an incredibly large amount of constraint on the system and will make it easier to analyse. Ben's First Obvious-to-me Claim: (1) In a sufficiently large, close community [s/close/(insular|incestuous|insectoid)/] the %age of person-person links which do *not* share any links via a third (non-person?) entity is zero. The interesting thing is what 'large' and 'close' mean here. My guess is 'bloody small' and 'actually quite sparse'. Ben's Second Obvious-to-me Claim: (2) The optimal size of an online cabal is an inverse power-law of the download size of the cabal (in some suitable units). The thing which interests me is that, if (2) is true (for some vaguely interesting groups), can we calculate the power involved (the critical exponent). If all of that is true, then whether a cabal has a maximal size or not depends on the value of the exponent. If it's > 1, then it does, but it might be hugely large. If it's > 2, then the cabal probably won't ever be found in the wild at a size very far from its maximal size. [If people want to see my reasoning, then asking me offline might be a good idea. Some keywords: "Universality", "scaling", "Kenneth Wilson", "criticality"] [De jargonised version of (2): The more traffic there is on a group, the more people will drop out. Under certain circumstances - probably most circumstances - we can find out how many people are likely to drop out at certain traffic rates. There might well be a maximum effective size for online groups, which we could determine by measuring things. The downside is, in practice, online groups which have been around for a while might always *be* at their optimal size] Ben
Re: networks and scalability
On Tue, Sep 24, 2002 at 11:49:48AM +0100, Simon Wistow wrote: > > This is something I've been looking at for some freelance consultancy > work on mobile communities. > > There are 4 interlinked problems : > > 2) Too much traffic and people get swamped and kind of just zone out and > skim read a lot of stuff. On a mobile device this swamping can happen > very quickly. People also don't like paying to download content they're > not going to read. > > 4) If you don't then people might get narked about paying to download > Arsenal gossip and getting Buffy spoilers. Plus it can act as an > inhibitor to people joining - they get there, take one look and scarper I don't think a per-piece model works for content. Take a look at the way The Economist web site handles premium content. The per piece model they have is pretty clearly designed to drive people towards a subscription. I would almost consider buying a subscription if their pro-war stance wasn't quite so repugnant. I'll settle for scarfing the free content off their website. Ben
Re: Good Training London?
On Tue, Sep 24, 2002 at 05:20:28PM +0200, Derek Revill wrote: > > Can anyone recomend a good all-round Perl training course or company in > London? I live in Amsterdam but want to come to london for training in > english. I'm pretty much at beginner level (or a bit beyond) and would like > to broaden my knowledge in all the basic areas (I do a lot of data > transformation and work with XML). I found some courses by 'gbdirect' - > anyone know them or of others? Recommendations appreciated. I've done some work for Learning Tree, and the instructors I'm aware of over there are very good in and of themselves. However they are very tied to a course structure and syllabus they have almost no control over. The material tends to be geared to the requirements of large corporates and mediocre students. If the 2000 quid price tag for a week doesn't seem like a lot of money, or if someone else is paying, I'd recommend them for a basic grounding. I may also know a small firm which might be interested - if you'd like to mail me some detailed requirements offlist. Ben
Re: [JOB] it's java though
On Mon, Sep 30, 2002 at 11:11:42PM +0100, Nicholas Clark wrote: > > I don't believe that web pages that have scripts > in them are good. > > Unless I can run the scripts in a sandbox environment. Which, to the best of > my knowledge I can't, and even then the presence of scripts in web pages > prevents me from programatically validating web pages. [I can't believe I'm about to defend this point of view, but there we are...] I was thinking about this just this morning, and I realised that scripting languages are pretty much essential on the client-side, if you're serious about producing something which is even vaguely usable as a real application. Hold on. I saw that jaw drop, in certain quarters. Let me explain. AFAIK, HTML is being used (wrongly, but there we are...) for two separate things: document markup and lowest-common denominator UI. The former, it's pretty shit at, except for text + a few pretty pictures. The latter, it's really, really shit at. It gets worse. If you care about correctness, adherence to standards, and just good old not having to error-correct every single thing you do, then You Must Use POST For Almost Everything.[1] Without a scripting language available on the client side, the only way to trigger a POST action from your UI is by a button push. Take a couple of minutes to think about how even a fairly simple call-centre application would look if every action had a and associated gubbins to go with it. Or a webmail application. Or a stock ticker. Or If all you want the web to be is a vast collection of same-y looking documents, then you can just about get by without clientside scripting, providing you're prepared to put up with curmudgeonly navigation. I'd like a bit more. Of course there are problems, not least of which is "wtf does a braille browser do with javascript", but then, I never said it wasn't shit. It's just what we got. Ben [1] There are, of course, security (eg X-site scripting, cookie stealing, etc) implications about GET as well, but I don't want to get into those here.
Re: similarity detection
On Tue, Oct 08, 2002 at 12:11:38PM +0100, Shevek wrote: > On Tue, 8 Oct 2002, alex wrote: > > > indeed - i seem to vaguely remember that i didn't use the sqrt in my > > postal sector[0] comparisons (it was to calculate nearest specsavers > > retail outlets to a postcode) and sql looked something like this: > > Metric space theory tells you that your distance computation is valid > whether you square or not. It's still a valid metric. The unit ball is a > slightly different shape ... Nonsense. d(x,y) = (x1 - y1)^2 + (x2 - y2)^2 + (x3 - y3)^2 and g(x,y) = sqrt d(x,y) have precisely the same unit ball. Ben
Re: similarity detection
On Tue, Oct 08, 2002 at 01:18:13PM +0100, Shevek wrote: > > Actually I was referring (with much gesticulation) to different metrics > having different balls, but ... yes, in this case you are right. > > The point being you don't need to square either, in which case you do get > a different ball. You mean: m(x,y) = |x1 - y1| + |x2 - y2| + |x3 - y3| the so-called Manhattan metric? This does have a different unit ball in a point-set sense, but please note that the point-set topology that these metrics infer is identical[0,1] Ben [0] In fact there's a theorem which proves that any metric on a finite-dimensional metric-space (ie, one which is a subset of R^n) produces the same open-set topology as any other. [1] Closed sets are problematic, because I can produce metric spaces which have holes in them, and sequences which *ought* to be convergent but converge to a point which ain't there
Re: applying patterns
On Tue, Oct 08, 2002 at 04:34:56PM +0100, Nigel Wetters wrote: > On Tue, 8 Oct 2002, Paul Johnson wrote: > > > And anyway, in my day "pattern" was spelt "recipe" :-) > > Sort of. Patterns are small ideas. They work well in combination. Many > recipes are larger, and solve whole problems. I particularly like the antipatterns book. I've felt that one could use patterns as a way of showing that you're doing the Right Thing, by stitching together small best practices. The great thing about Antipatterns is that they remind you that you can still stitch together a big load of code that's fine at the small scale but turns into an ungodly monstrosity at the level of a whole system. I also found myself reading the descriptions of the Antipatterns and going "Seen that one. Oooh, and that one. Wow, I've actually heard a cow-orker given exactly that spiel to describe the project." Happy days. Ben
Re: More books
On Mon, Oct 14, 2002 at 12:27:57PM +0100, Simon Wistow wrote: > On Mon, Oct 14, 2002 at 11:24:09AM +0100, David Cantrell said: > > Who wants to review Graphics Programming with Perl? > > Me! I tech reviewed this book so I'd be interested to see if they solved > some of the problems. Bah! You beat me to it. Ben
Re: Next Thursday; was: small hairy Belfast.pm geek...
On Fri, Oct 18, 2002 at 10:26:57AM +0100, Nicholas Clark wrote: > > However, it is a moderately long walk from the nearest underground station. > (it's managed to pick a sweet spot (a bitter spot?) moderately equidistant > from three underground stations, so it's certainly an above average > distance) But, really it's not *that* far from the tube. Nowhere which is still in Zone 1 is *that* far from a tube station. > It is approaching one of the less nice parts of central London, sufficient > for one london.pm member's wife to insist that he doesn't go there. I have to take issue with this somewhat. London may not be a land of fairies and elves, but it is for the most part perfectly safe. > That may sound trivial - in some ways it is - 1 person out of many, > subjective rather than objective. But it's also the first location (that > I've been aware of) where this has happened, and I admit that I don't find > it a particularly pleasant walk late at night down Gray's Inn Road. In which case, may I suggest Oxford St or Dean St after about 2230. This should, of course, be viewed as entirely subjective, as I can be famously thick-skinned and regularly go drinking/clubbing in Kings Cross on a Friday night. (And I've never had a problem. *touch wood*). Ben
Re: ADSL again
On Fri, Oct 18, 2002 at 03:09:18PM +0100, Simon Wistow wrote: > I want to get ADSL. > > The question is do I go for a wires only option and buy my own ADSL > modem or do i go for Nildram's managed USB frog at a 25 quid one off > charge and then 8 quid a month? Wires only. I have yet to be convinced that anyone reading this list would be better off with a managed router. I take some convincing that *anyone* can't handle a wires-only connection. Even my adequate-for-me Dlink was insanely easy to set up. I'll even come round and help you set up wires-only in the alternate universe where it gives you any problems. Ben
Re: ADSL again
On Fri, Oct 18, 2002 at 03:25:06PM +0100, Leon Brocard wrote: > Simon Wistow sent the following bits through the ether: > > > I want to get ADSL. I've pretty much decided on Nildram, specifically > > Home 500 Lite : > > I have this with a nice D-Link 504 4-port ADSL Router in my flat and > am setting up the same thing at parent's as it's so good and just > works. I highly recommend this. The Dlink is a piece of piss to set up. Be aware, however, that it isn't the best router in the world, and not everyone likes it. It's perfectly adequate for my needs, however. Ben
Re: ADSL again
On Fri, Oct 18, 2002 at 03:56:56PM +0100, Nicholas Clark wrote: > > So your advice for anyone contemplating ADSL would be an architecture > something like > > BT > | > | > DSL router/hub (such as Dlink) > | > Firewall box (eg cheap x86 running some sort of BSD or Linux) > | > Hub > / | \ > Stuff > > If the firewall box has (or can be fitted with) several internal interfaces, > is it a viable, cheap and secure system to also use it as a hub? This could be quite a good idea for, eg, people who wanted to provide open wireless access - the wireless access device traffic can then be kept separate from the internal networks. This probably comes under the heading of 'Well, duh', but it's Friday afternoon, so I think I'm allowed some latitude. Ben
Re: Writing XML from Perl
On Thu, Oct 24, 2002 at 05:16:13PM +0100, Dominic Mitchell wrote: > Marty Pauley wrote: > > On Thu Oct 24 15:13:09 2002, Kris Boulez wrote: > > > >>What is peoples preferred module for writing XML (files) from within > >>Perl ? The data is read by a perl script from different (heterogeneous) > >>sources and stored internally in a tree of objects. > > > > > > I use XML::DOM for that sort of thing. I haven't tried anything else > > since XML::DOM has worked well. As I walk the tree each node calls > > createElement for itself then appendChild for any branches. > > XML::LibXML (as mentioned in another post) is likely to be a lot > quicker, as it's based on a C library. XML::GDOM has a partially or totally SAX2 compatability, is based on libxml2 and has a virtually drop-in replacement interface for XML::DOM. It was what I used the last time I did any XML. But then, I hate XML. Ben
Re: Apache::DProf
On Thu, Oct 17, 2002 at 03:58:22PM +0100, Barbie wrote: > I'm trying to get Apache::DProf running with mod_perl. Unfortunately the > resulting tmon.out files all contain basic header info and a single line, > such as "@ 2 0 147044", under PART2, which corresponds to the rrun variable > settings. I must be missing something, but the perldoc isn't very helpful, > as there's more about how to configure Devel::DProf under mod_perl, than > there is of Apache::DProf itself. I had enormous problems with Apache::DProf - it would inevitably produce truncated/corrupt tmon.out files - which looked to my untutored eye as though something internally was getting confused as to where in the call stack it was, and ended up trying to pop an empty stack at context exit. Some dark nargery in a really ill-advised fashion seemed to confirm this, but I lacked the skill to fix it, and started using Devel::Profiler::Apache instead. I can explain further if you'd like, and please let me know if you get anywhere with Apache::DProf. Ben
Re: Apache::DProf
On Thu, Oct 17, 2002 at 05:08:13PM +0100, Barbie wrote: > From: "Ben" <[EMAIL PROTECTED]> > > > ... started using Devel::Profiler::Apache instead. > > Tried that and it has the same results, all the tmon.out files just contain > a single line below the PART2 line :( Running dprofpp -u just returns to the > command line, with nothing being printed. Anything odd about your mod_perl, Apache or perl builds? Ben
Re: webmail
On Tue, Oct 29, 2002 at 09:57:48AM +, Alex McLintock wrote: > At 09:29 29/10/02, the hatter wrote: > >I wish people would stop writing tied-in templating systems, and define a > >meta-templating system so apps could easily be moved between templating > >systems. > > I'd say that XML and XSLT fits the bill. And I'd say bwahahahaha. Seriously, though. I've experienced such enormous slowness and lack of scalability with XSLT that I couldn't really consider using it for production systems at the moment. Call me old school, but I think I want a thing which is doing some sort of transformation to some sort of output stream to, well, be written in a proper programming language, rather than go through all the hoops and massive verboseness of something like XSLT. I am a bit of an XML bigot, though. However, I've just started looking at XPFE and Mozillas framework stuff, and despite a few minor annoyances so far, it looks like it might help cure my XMLophobia a bit. Ben
Re: Usernames?
On Tue, Nov 05, 2002 at 06:24:31PM -0500, Adam Turoff wrote: > On Tue, Nov 05, 2002 at 10:24:42PM +, Paul Makepeace wrote: > > The traditional restrictions on web usernames are things like only > > alphanumerics, and usually lowercase to reduce user confusion/burden > > remembering. > > > > Why not allow (embedded) whitespace, punctuation, and so on? > > Characters like . cause problems because they've always been illegal. > For example, chown accepts a syntax of 'chown user.group' because . > is illegal. So chown breaks when you have a username like s.avery > or m.thibaut. > > The shell has certain expectations about when and where # will be used, too. Err, these only seem to apply to situations where a web user maps down to an underlying *nix user. FWIW, I'd never allow users of any system anything other than whatever is considered current for *nix login names, just because of the possibility (read: likelihood) that my system would get maintained by someone who I wouldn't trust to Do Things Properly wrt char escapes, etc. Ben
Re: Usernames?
On Wed, Nov 06, 2002 at 12:47:09PM +, Paul Makepeace wrote: > On Wed, Nov 06, 2002 at 11:13:43AM +, David Cantrell wrote: > > Stupid programmers forgetting to appropriately quote/escape data before > > stuffing it into a database. > > > > INSERT INTO users (userid, username, passwd) VALUES > > (usersequence.nextval, '$username', '$passwdmd5hash') > > Why would anyone not use bound variables? I've seen quite a few who honestly believed that using explicit values was faster, because in their minds it must be easier for the DB to process. Not only are they emphatically *not* easier or faster, but some databases (notably Oracle) can suffer serious performance problems because of this. "Always use bind variables" seems to me to be one of those things that should be included in every basic reference of how to interface with an SQL database. It's been sadly lacking in a couple I've seen recently, however. It was a carelessness I occasionally succumbed to myself, but then I saw how much impact it could have on a Oracle installation. Non-bind SQL: Just Say No. Ben
Re: Usernames?
On Wed, Nov 06, 2002 at 01:20:12PM +, Mark Fowler wrote: > On Wed, 6 Nov 2002, Ben wrote: > > > Non-bind SQL: Just Say No. > > And that would make an excellent lightening talk. Anyone (who's not > already doing one, this includes Ben) want to give it? Errr. I thought I was doing 'When Not to Use mod_perl'. I have a smallish chunk of AxPoint assembled along those lines Ben (cc: to london.pm in case direct mail is somehow not getting to you - I definitely confirmed this last week or the week before...)
Re: Apache question
On Wed, Nov 13, 2002 at 03:14:42PM +, Kate L Pugh wrote: > > Can I make apache proxying and server-side includes play nicely together? > > I have a lightweight server serving static content and server-parsed > files, and proxying dynamic content through to a mod_perl server. > Unfortunately the server-parsed pages have '#exec cgi' directives in > them, and the mod_perl server just isn't seeing those requests. Am I > trying to do something impossible? > > When I enable ExecCGI on the lightweight server everything works fine > -- but I don't want that server to be serving out cgi. And I'm sure > my rewrite rules are OK because when I turn ExecCGI off again on the > lightweight server and call the things that are 'exec cgi'ed directly, > the mod_perl server sees and serves them up fine. > > The interesting parts of the relevant config files are: > > -- > apache_plain.conf: > > # proxying to the mod_perl server > ProxyRequests off > ProxyPassReverse /cgi-bin/ http://localhost:82/cgi-bin/ > RewriteEngine on > # make sure that static and SSI content is served by plain server > RewriteRule \.(gif|jpg|html|shtml|css|txt)$ - [L] > RewriteRule ^/(.*\.)(cgi|pl)$ http://localhost:82/$1$2 [P] > > -- > apache_perl_ukps.conf: > > PerlModule Apache::DBI > PerlModule DBI > > > SetHandler perl-script > PerlHandler Apache::PerlRun > Options +ExecCGI > PerlSendHeader On > > > Can anyone help? OK. I'm not that much of an expert, but I'll have a go: The relevant thing to look at is the include_cgi function of mod_include.c It works by triggering a GET subrequest for the relevant CGI include. (Have a look at ap_sub_req_lookup_uri() in http_request.c) So what we need to do is persuade Apache that this subrequest (which is still in the URI space of the local server) really wants to be subjected to the same rules (ie the proxy stuff) as a genuine request. Now, IIRC, there *is* a way to do this - for much the same reasons that per-directory configuration for mod_rewrite works. I'll have a dig about in the mod_proxy documentation and see how/if subrequests are made to behave like real requests. Ben
Re: Announce: Devel::LeakTrace
On Tue, Nov 12, 2002 at 04:50:50PM +, Richard Clamp wrote: > I just allowed this to escape to CPAN, more information in my use.perl > journal, and of course the original source. I hacked this slightly to use stderr instead of stdout, in an attempt to make it play nicely with mod_perl, but I'm getting this error: Can't locate auto/Devel/LeakTrace/show_used.al in @INC (@INC contains: INC path> ) at /usr/lib/perl5/site_perl/5.6.0/i386-linux/Devel/LeakTrace.pm line 14 END failed--call queue aborted. It works fine if called from a normal script. Any ideas? Ben
Re: Technical Reviewer
On Thu, Nov 14, 2002 at 09:34:06AM +, Alex McLintock wrote: > Manning have asked me to be a technical reviewer for one of their (non > perl) books. Yay. Which one? > Is it standard practice that technical reviewers don't get paid? I am happy > to do the review anyway because I run DiverseBooks.com but I'd like to know > that I am not being taken advantage of. I was paid by O'Reilly for one of the books I did for them (but not for the other). It wasn't a lot, but it did also involve free books, ego strokes, etc. As I understand it, though, it rather depends on the author, and the stage at which people get involved in the process. I have a friend who has done tech reviewing for Manning, so I'll ask him, but IIRC he wasn't paid. Ben
Re: Perl and CC processing
On Fri, Nov 15, 2002 at 03:32:45PM +, Aaron Trevena wrote: > I have been looking at doing my own hosting, etc and was wondering what > people reccomened for CC processing. > > I've heard of a couple of big names like WorldPay, and TrustCommerce has a > nice offer but is US-oriented. > > datacash looks good - uk based and perl api. I would not consider using datacash. Their reliability and technical quality are not good, and from what I've been told about their backend architecture... Ben
Re: contracts
On Tue, Nov 19, 2002 at 08:08:26PM +, Graham Barr wrote: > > For a while now I have been doing small jobs on the side, while > being employed. But I have been thinking about leaving my full time > job and doing contracting. Now before you all tell me I am crazy, > I have my reasons for wanting to leave :) > > But the question I have is, which is the best route. Is it better > to start a ltd company or just go self employed. I have read a lot > of online info fro various web sites, but none I found give real > pros and cons either way. Ltd company. Definitely. Apart from anything else, some companies IME won't employ contractors who don't have a ltd company. Something about certain types of corporate insurance. Ltd companies are cheap (70 quid, if that) to set up, easy to administer and have advantages from the point of view of liability and tax. The only downside that I've found is that an accountant is pretty much essential. But again, they aren't expensive and seem to be worth the cash I'm shelling out on them[1]. Ben [1] And I didn't have to look very far to find one. :)
Re: RMS seminars in London
On Fri, Nov 22, 2002 at 10:00:33PM +, Earle Martin wrote: > On Thu, Nov 21, 2002 at 12:10:28PM -, B. N. Moran wrote: > > People might like to know that Richard Stallman will be speaking at the > > London School of Economics on 2nd and 3rd of December. > > I'm going to both; anyone else? Pub beforehand perhaps? I've got tickets for the Tuesday one. I'd be up for meeting beforehand, and certainly pub after. Ben
Re: Apache2
On Tue, Nov 26, 2002 at 11:11:14AM +, Jonathan Peterson wrote: > > I'm rather unimpressed by the complete > absence of seemless backward compatability, but I suppose change is good. Errr, I'm not. The 1.3 Apache API is badly geb0rken in a couple of places. Fixing that, and doing the correct thing instead is far more important than being backward (compatible)[1], IMO. > *I still think building a working mod_perl is one of the hardest > software install / configuration jobs going. apxs2 helps somewhat, but > it's still insanely hard. On which platform? It varies significantly between Linux, Solaris, Digital and *BSD[2] and can be sensitively dependent on quite surprising things. It does require a bit of thought and should not be plowed into blindly, and is generally harder than it needs to be. Perhaps I should give a rant about all this stuff. Oh, wait Ben [1] I've quite often felt the second word of backward compatible was unnecessary. [2] Which are the only platforms I've built it on.
glaziers
Hi, (Not that I use london.pm as a Source Of All Knowledge or anything) I'm in need of a glazier to repair my living room windows that were damaged in the recent-ish storms (yes, I really am that lazy). Can anyone recommend someone who's reliable, works in London and not stupidly expensive? Thanks, Ben
Re: Search and Extract
On Thu, Dec 05, 2002 at 12:02:43PM +, Chris Ball wrote: > >> On 5 Dec 2002 11:02:41, Julian Martin <[EMAIL PROTECTED]> said: > >> I would like to search some html pages for a keyword and then >> extract the blah, blah..keywordblah and then >> put the blah, blah..keywordblah's into a >> results page. Any pointers would be great ! I have Perl cookbook >> for any reference but cannot find something like this in it. > > In case you were wondering about it, it's usually a very bad idea to try > and use regular expressions instead of an HTML parser for this sort of > thing. This is entirely true, but I thought it warranted a bit of explanation as to *why* it's a bad idea. Sorry if this is a bit basic and/or patronising, Julian - it isn't intended to be. Parsing any sort of data can be surprisingly tricky. HTML is an especially nasty case because it's sort-of human readable, and humans are messy creatures and tend to produce ad-hoc and sloppy representations of data. Not only that, but HTML tools have traditionally produced bad HTML, because browsers have always tried to be accomodating and have a good stab at displaying what they think the author intended. Some of them do a very good job, but that doesn't encourage either HTMLers or tool makers to produce HTML which actually does what it's supposed to. To add to the nightmare, the HTML standard, such as it is, has evolved over many years. Earlier versions are still supported, and there's a lot of valid once-upon-a-time HTML out there. While everything might well work fine with the dataset you currently have, when you come to extend it (as will happen - Murphy's Law dictates that it must be so) - things will break. You'll start running into cases you haven't though of, and sooner or later the nasty mind-fucking edge cases which can drive you into Lovecraftian madness and the Lands Beyond will rear their ugly heads. By this time, your code will be so tortuous, from all the modifications you've had to make along the way, that you can barely understand it yourself any more. Fortunately, it's not all bad news. There are some very nice people out there who have already braved the nastiness, and they've made a module of code which will take all the pain away. Or a good chunk of it, anyway. They've put it up on CPAN and it's all yours. Using it really would be a good idea. Trust me. Ben
Re: UML diagramming tool
On Fri, Dec 06, 2002 at 11:03:54AM +, Mark Fowler wrote: > Does the class have any consensus on UML diagramming tools? In > particular we're looking at Visio (primarily as we can use it for doing > other things too, and it'll run under Windows and Linux[1].) Apart from > it being from Microsoft, does anyone have any thoughts? We're (obviously) > going to be primarily documenting our Perl code, and aren't really > interested in code generation or diagram extraction technologies. I've never been too sure of the approach of "drawing pictures to document code". UML does at least recognise that you need more than one set of pictures, and their relationship to each other is a bit tatty and non-orthogonal, but still. > I've not touched UML since university so I admit I haven't much recent > experience. Are there other good *mature* tools that will run on multiple > platforms, that are around the same price range as Visio? I'm guessing > there might be something in Java that could help us out here. If you *really* don't want code generation or extraction, then a good diagram drawing package should do. I've always liked dia, on the rare occasions I've had to have anything to do with UML. > One thing we think might be a problem with Visio is it doesn't seem easy > to hide parts of the diagram that you're not concerned with at the time. > Is this going to be a problem? I would recommend that you read: http://www-inst.eecs.berkeley.edu/~maratb/readings/NoSilverBullet.html It may be old, and a bit curmudgeonly, but there are some excellent points about the non-geometrical nature of code. In addition, there is some good stuff in The Humane Interface by Jeff Raskin, about the nature of zooming interfaces, which have always seemed to me to be necessary to provide even a limited representation of software. Even though any sort of graph of software worth drawing would undoubtedly be non-planar, and all of Raskins remarks as presented apply only to the 2d case. Congratulations, btw. Ben
[JOB] High level real job available
Hi, http://jobs.perl.org/job/533 I know for a fact that this one is real, it's all perl work, *but* you will need to be very, very good. Sysadmin experience is pretty much essential, as is running herd on a team, and looking after the production kit. It's also quite stressful, and you'll need to do some back-line support. If anyone is interested, please mail me offlist, as I'll be able to brief you further and introduce you to the agent if you wish. Ben
Re: Estimating accuracy of IP->country lookup
On Tue, Jan 07, 2003 at 09:50:15AM +, Nigel Wetters wrote: > It would be useful for those using the IP::Country::Fast module in log > analysis to know the accuracy of the result. At this point, I don't have > a clear idea of how to make this estimate. > > For example, it is possible that multinational companies (and ISPs) may > use US-registered IP addresses outside of the USA. It is difficult to > see how this could be measured. Um. If you can get hold of the underlying data, rather than just the press release, then the data referred to in: http://www.linx.org/press/releases/081.thtml may be helpful, if you make some possibly sane assumptions about how LINX routing works. I have half an idea about how to do this, and know a couple of people who should be able to justify or shoot down my assumptions. > Greg suggested on IRC that it might be an idea to ask people their IP > addresses and country. Does anyone have any suggestions on how to > automate the process on a large enough scale to get a significant > result? That will certainly help once the sample size is sufficiently large enough. I'm quite interested in this as a problem, feel free to mail me offlist if you want to chat about it and can put up with my possibly not-sane whiffling. Ben
Re: [JOB] Perl/Oracle/Linux team leader
On Wed, Jan 08, 2003 at 01:15:06PM +, Adam Spiers wrote: > This is genuine and hot off the press: > > Perl / Linux / Oracle looking at a Team Leader type role - slightly > longer term as a form of CTO (Chief Tech Officer) - still very much > hands on though. This is the same as a lead I posted before Xmas. I am informed that it's a bit naughty of the agents to be releasing this much detail so early in the process. The job, however, is real. Ben
Re: [JOB] Perl/Oracle/Linux team leader
On Wed, Jan 08, 2003 at 03:13:59PM +, Nicholas Clark wrote: > On Wed, Jan 08, 2003 at 03:02:49PM +0000, Ben wrote: > > On Wed, Jan 08, 2003 at 01:15:06PM +, Adam Spiers wrote: > > > This is genuine and hot off the press: > > > > > > Perl / Linux / Oracle looking at a Team Leader type role - slightly > > > longer term as a form of CTO (Chief Tech Officer) - still very much > > > hands on though. > > > > This is the same as a lead I posted before Xmas. I am informed that it's > > a bit naughty of the agents to be releasing this much detail so early > > in the process. The job, however, is real. > > I thought I'd seen it before. Any clues (from insiders, or just plain > guesses) as to why it's being advertised again, and seemingly touted as > "new"? Particulary as there seem to be enough CFT-enabled perl mongers to > fill it several times. It's not being advertised again. The job is still at a pre-interview stage, and likely to remain so for at least a few more weeks. It hasn't formally been advertised yet at all, and no-one has been interviewed for it. Also, it's worth making the point that this isn't a perl code-cutters role. There's a lot more to it than that, and I think it's going to be a very difficult role to fill. Not because they want the moon and are prepared to pay sixpence, but because it's genuinely a bit different and quite stressful. Ben
Re: [JOB] Perl/Oracle/Linux team leader
On Wed, Jan 08, 2003 at 04:45:20PM +, Adam Spiers wrote: > Nicholas Clark ([EMAIL PROTECTED]) wrote: > > On Wed, Jan 08, 2003 at 03:02:49PM +, Ben wrote: > > > On Wed, Jan 08, 2003 at 01:15:06PM +, Adam Spiers wrote: > > > > This is genuine and hot off the press: > > > > > > This is the same as a lead I posted before Xmas. I am informed that it's > > > a bit naughty of the agents to be releasing this much detail so early > > > in the process. The job, however, is real. > > Hmm. Maybe Jeff didn't want me to disclose that much info then. But > it all seems rather silly: > > a) Why would you ever want to release details in a trickle rather > than just getting the spec out there? Errr. How about precisely to avoid pissing off the qualified candidates by releasing loads of details for a job which isn't at the interview stage yet? No-one likes to be kept hanging about by an agent who seemingly can't deliver. OTOH, if I was advertising a job, I'd want to put some feelers out, see what sort of folk were available, and get a feel for the value the market would put on a job, etc. TBH, I strongly suspect that this a bit of a mountain out of a molehill, and all that's happened is the agent got a bit over-zealous. Ben
Re: [ANNOUNCE] Tech Meet Thurs 23rd Jan @ Yahoo! (+ tonight, social meeting)
On Thu, Jan 09, 2003 at 03:06:36PM +, Mark Fowler wrote: > On Thu, 9 Jan 2003, Mark Fowler wrote: > > > Tech Meeting 23rd Jan > > It's been pointed out to me that I should have mentioned the time. The > Tech meet will open it's doors at 6:45pm for a 7:15pm start, meaning we > should be leaving the building sometime between nine and half nine. > > Right, what else have I forgotten? The before / after pub? Ben
Re: Ski Social!
On Fri, Jan 10, 2003 at 10:07:08AM +, Joel Bernstein wrote: > When the lovely snow was coming down, Robin Szemeti and I discussed how > we should do a London.PM-On-The-Piste social this winter. I was hoping > to gauge opinion on this... Would anybody be interested in a weekend in > the Alps skiing or snowboarding? I'm in. The weekend of the 14th is bad for me, as I'm in NY to see the missus. It might be as well to avoid the weekend of the 8th-9th as well. Might I tentatively suggest the last weekend in Feb? Could I also suggest a high resort, to avoid problems with lack of snow - especially for beginners? One resort I'm particularly fond of is Saas-Fee, in Switzerland, http://www.saas-fee.ch/index-e.htm It's a very friendly, and family resort, and was the resort used in Wham's Last Xmas video, for the pop-pickers among us. The only drawbacks are that it's a 3hr train, then a 1 hr bus from either Geneva and Zurich. Flights are definitely available from LTN, LHR and LGW, but obviously City would be easier for most people. This might be a bit of a long journey for just a weekend, but the resort really is very nice. I spent 2001 there and it was great. Anyway, enough of my rambling. Anyone else got a resort in mind? Ben
Re: copyright and NFS
On Mon, Jan 13, 2003 at 11:32:38AM -0600, Nigel Hamilton wrote: > > Soon enough the 'rage against the machine' rhetoric is going to > sound tired ... both Lessig and Stallman, can choose to send a more > positive message (albeit a bit boring) ... the fact is no corporation ever > had an 'idea' ... 'ideas' are the property of individuals! And > intellectual property is about protecting them. > > It's what we do with IP rights that's important ... part of the > problem is letting authors know that they have power and that their > intellectual efforts can/should be rewarded! By IP here, I take it you mean copyright law. I can't see a way that patent law would be particularly relevant to this discussion. Ben
Re: Potential new module
On Tue, Jan 14, 2003 at 10:26:40AM -, Ivor Williams wrote: > I've been reviewing what I have learned from the LWP stuff I have been doing > for Grubstreet. This involved form filling and POST method. > > I was struck with the feeling that boiling down the form data is something that > has probably been done many times over - but a search didn't find anything > obvious. Yes, I'm sure it has been. That doesn't mean that anyone ever released it to CPAN, though. :) Between us, london.pm must have written $LARGE_NUMBER of templating systems, but only a couple actually made it out. Or how many reasonably generic CGI-as-dispatch systems were written back in the heady days of dotComNonsense? > >From the existing code that I have written, is a sub formdata, which takes the > HTML page and form name as parameters. The form name is optional; if no form > name is specified, the routine picks up the first form on the page. > > The sub returns a list of key/value pairs. Thinking about it, I realised that > if the calling code turns it into a hash, this could lose any duplicate keys. Except for checkboxen (and possibly even then), duplicate keys on forms are a very bad idea. One way I've seen them go horribly wrong is with a couple of URL-compression / cookie-caching proxies as used on The Portal Which Shall Not Be Named. The idea of these doohickeys is to improve access for phones, and other brain-dead clients, by shortening the (RFC-breaking) huge GET URLs that portals always seem to end up with, and by keeping cookies on the server side. By this stage, I'm sure people have guessed how good the hashing and storage of these little sweethearts is. As soon as you start pointing them at forms with duplicate keys, they start dropping data on the floor. The other thing I always try and keep in mind is nested forms. They are a very good way to see how sturdy your application or web-snarfer really is. They're invalid, nasty, rather more common than people seem to realise and have a horrible habit of making otherwise well-behaved web-tech spazz out. I've even seen a (Java) example where if you parse a document containing nested forms as XML, all appears to be well, and then if you cast it back to an HTML-type object, the JVM crashes. Luverly. > At this point, the light of recognition came on in my mind. This was a very > familiar concept, that of a CGI object. > I could make formdata return a CGI object or something inheriting from CGI, > giving access to all the input fields via $form->param. Besides being capable > of being submitted via a normal POST of encoding type > application/x-www-form-urlencoded, I would also like the code to be able to > handle file uploads and encoding type multipart/form-data. Please don't make it return a CGI object. CGI.pm is a nasty, bloated piece of crap which should have been retired years ago. It uses what is now very non-standard technology, it leaks memory under mod_perl and in many ways is an example of how not to build a module. Ben
Re: Ski Social!
On Fri, Jan 10, 2003 at 06:25:17PM +, robin szemeti wrote: > > > Absolutely. I'm a huge fan of Val d'Isere, where I worked for a season. > > Chamonix isn't too convenient (valley town) but has marvellous skiing > > and is an hour from Geneva. > > I *love* chamonix ... so that would suit me. 5 seperate ski areas accessible > from the village, an hour or so from GVA on the train ... and theres the > 'vallee blanche' 22km off piste run if anyone fancies it :) add in decent > night life and being a real town rather than a ski resort, we might find it > easier to get 'long weekend' accomodation as the resorts are really just > geared up for whole-week packages. > > But, I'll ski anywhere for a long weekend, no worries. OK. No-one's posted about this for almost a week, so I thought I'd reopen the discussion. I think we need to decide: Who? Where? When? How Long? Then we can go and book it. :) I'm not electing myself cat herder or anything, I'd just like this to happen, so I propose that anyone who's up for it follows up to this mail, then we take it to private mail to avoid cluttering the main list, unless everyone else doesn't mind wuffle about skiing. Summary of my recollections: Who:Ben, robin, Joel, ... I know there were others? When: Last week in Feb? Or some other time. Trying to avoid FOSDEM and Valentines Where: Unknown. Val d'Isere, Chamonix and Saas Fee have been mentioned as possible. More ideas? "Somewhere high up with mountains"? Time: "A long weekend". Fly out Thursday night, take friday off, fly back sunday night? Other ideas? Ben
Re: tech meet?
On Sat, Jan 18, 2003 at 01:52:06PM +, Dave Hodgkinson wrote: > Is anyone interested in my hypothetical "Using Apache::Template in a PHP > stylee" talk? yes. Ben
Re: CVS Client
On Wed, Jan 22, 2003 at 03:03:23PM -, Neil Fryer wrote: > > Can anyone recommend a decent free CVS client, for W2K? To connect to a > Linux CVS server? The last time I had occasion to do this, I used a package called TortoiseCVS (http://www.tortoisecvs.org/) which made CVS much more like a Windows file browser. The non-techies in the company came to quite like it. At that time, the interface for WinCVS was not good. It may have improved since then. I certainly hope it has. Ben
[JOB] A few outstanding individuals needed
Hi, In the pub last night, a friend told me of a company who are actually hiring at the moment. They want Java for preference, but I was told that they do hire 'bloody clever people' who don't have Java. If you're CFT-enabled or just in the market for a new job, send me your CV and I'll forward it. I will receive a finders fee if you're hired, a percentage of which I shall donate to the charity of your choice if you wish. Or we could just spend it on beer. [Mark - I hope this is OK to post, please slap me if not] Ben
Re: [JOB] A few outstanding individuals needed
On Sun, Jan 26, 2003 at 05:31:20AM +, Piers Cawley wrote: > Ben <[EMAIL PROTECTED]> writes: > > In the pub last night, a friend told me of a company who are > > actually hiring at the moment. They want Java for preference, but I > > was told that they do hire 'bloody clever people' who don't have > > Java. > > > > If you're CFT-enabled or just in the market for a new job, send me > > your CV and I'll forward it. > > > > I will receive a finders fee if you're hired, a percentage of which > > I shall donate to the charity of your choice if you wish. Or we > > could just spend it on beer. > > Whereabouts (don't need the company name)? What kind of pay? What kind > of work? London Zone 1, 'good money', which from the friend in question means certainly above market rate and it's all development work. I was told the most important things were people being very smart, but also that they be compatible with the prevailing corporate culture. No, I don't know what he means either. I'd be dubious about the whole thing, but my friend is a designer, not a recruitment bunny, and I know that he's got jobs for friends of mine before, so... I'll try and get some more details, and post them here if I get them. Ben
Re: [PUB] Fitzroy Tavern, Fitzrovia
On Mon, Jan 27, 2003 at 11:45:59AM +, Joel Bernstein wrote: > On Mon, Jan 27, 2003 at 11:37:57AM +, Kate L Pugh wrote: > > A suggestion from Ben - the Fitzroy Tavern. It's another Sam Smith's > > pub, but Ben says the beer is decent, so I'll leave him to defend that > > point of view. It's on Charlotte Street, parallel to Tottenham Court Road: > > > > http://www.streetmap.co.uk/streetmap.dll?G2M?X=529522&Y=181615&A=Y&Z=1 > > > > and it's kinda famous: > > > > http://www.professorharbottle.co.uk/pub/londonwestend/fitzroy.html > > http://www.pennies-from-heaven.org/ > > > > (also > > >http://www.museum-london.org.uk/MOLsite/exhibits/creative/artistloc/1920/1920_fitzroy.html > > but I can't get to it at the moment) > > > > I phoned them and they do have a function room, in the basement, with > > toilets on the same level, but it costs 25 quid to book (and another > > 25 quid if we want the bar open) and the person I spoke to sounded > > dubious about us getting more than 35 people in there. They do a full > > menu until 9:30pm. > > > > Is this worth looking at? > > No! No! A thousand times no! The Fitzroy Tavern is /horrible/ - rude > australian barstaff, poor beer, noisy, intrusively flashing fruit machines, > full of people in reebok classics trainers - I can't think of a less > appealing pub to go to. You're obviously on the other side of the branch cut from me. I've been drinking in there since about 1995, and I like it. It can get a bit packed, so a meet where we didn't have the function room (and bar) probably would be less than optimal. The OB is reasonably well-kept, though not excellently so. Horses for courses, I suppose. Ben
Re: [PUB] Spread Eagle, NW1
On Tue, Jan 28, 2003 at 09:22:39AM +, Joel Bernstein wrote: > > The pub is on the corner of Parkway and Albert Rd, about 4 or 5 minutes walk > from the tube. Plenty of seating, friendly bar-staff. I saw 3 or 4 different > lagers, Guinness, 3 or 4 bitters on tap, and of course Double Chocolate > Stout. And, of course, a london.pm'er behind the bar! > I don't think it would be suitable for a pubmeet (possibly too far North, no > food, didn't see a function room, and the toilets are downstairs (not good > for disabled people) but I recommend it as a nice place to drink in Camden. I agree. It also can get very full. Camden is probably too far North for a pubmeet. Which is a shame, as there are some good pubs outside the Circle Line. Ben
Baby Jesus is crying...
Just read it and weap. http://www.datapower.com/products/xa35.html Ben
dragons
Is anyone thinking of going to this tonight? http://www.tateandegglive.com/event1_cai.html Ben
LibXML segfaults
I was writing a gremlin with WWW::Mechanize, and in doing so realised I didn't grok XPath as well as I thought I did (has someone got some pointers for me?) Then I discovered this: The value method of XML::LibXML::Attr appears to trigger a segfault if it's passed a value (I hadn't read the documentation and so didn't know that I couldn't treat it as a hash lookup, but have to treat it as an Iterator). So I wrote a test script: #!/usr/bin/perl use strict; use warnings; use XML::LibXML; my $b = 1982; my $a = \$b; bless $a, 'XML::LibXML::Attr'; 1; [bene@green gremlins]$ perl test1.pl Segmentation fault Anyone have any ideas? I know I'm being a bit naughty by blessing $a into someone else's package, but I don't think a segmentation fault is entirely the correct thing to do. Ben
Re: WWW::Map::UK::Streetmap - A tale of woe
On Mon, Feb 10, 2003 at 08:16:53PM +, Paul Makepeace wrote: > > Programmer cluefulness being equal, when did interpreted, profiled > languages start even slightly approaching the speed of compiled, > profiled languages like C(++)? So, in short, if we assume some things which aren't true, then we can infer some other things which aren't true? It's well-known that perl weighs the same as a duck, anyway. Ben
Re: plumbers
On Tue, Feb 11, 2003 at 11:49:25AM +, Dirk Koopman wrote: > > And just maybe I might retrain as one seeing some of the salaries / fees > plumbers are getting these days... To say nothing of all the kinky sex. Ben
Re: perl website on CD
On Tue, Feb 11, 2003 at 03:04:29PM +, nemesis wrote: > Randal L. Schwartz wrote: > > > > Second, why anyone still uses MySQL on a new installation now that > > PostgreSQL has surpassed it and gone on to being a "baby Oracle", > > while MySQL plays "catch up", I can't figure out. > > I heard/read somewhere that although MySQL lacks the features of PostgreSQL, it > is just faster for simpler operations. Maybe it isn't. It used to be. MySQL was so fast historically, because it wasn't really a database (Real Programmers Don't Need Referential Integrity and all that nonsense) but rather an SQL interface to a flat file. When the MySQL people started to realise that they needed to add real database features, the gap between PostgreSQL and MySQL narrowed. The last time I looked at it, PG was a little faster than My, but there wasn't a lot in it. Given the lack of speed difference, PG is infinitely preferable to My, at least in my little world. Ben
Re: Language Gentlemen and Ladies
On Wed, Feb 12, 2003 at 12:23:46AM +, Lusercop wrote: > On Tue, Feb 11, 2003 at 06:42:30PM +, David Cantrell wrote: > > https proxy? yay for man-in-the-middle fun! > > OK, I'll bite, given TLSv1 or SSLv3, what's the attack (bear in mind the > proxy is a "CONNECT" proxy)? I'm sure you can get a paper out of it. While we're on the subject of CONNECT proxies, it's worth telling people that there are some proxies out there which are broken and don't allow port numbers as part of the CONNECT line. For this reason, I tend to enable SSL on port 80 as well as 443 on my web servers so that I'm not denying anyone access to my SSL. There may be some difference of opinion about whether it is better to try and 'gently re-educate' users and producers of such software. However, I'm leaving for NY in two hours, so I won't speculate. I also need to pack. Ben
Re: YAPC::Europe & War
On Thu, Feb 13, 2003 at 01:03:38PM +, Simon Wilcox wrote: > > OK then, as we rot from the inside in the dust from the dirty bombs let > off around London. Better ? Where does one buy an Isle-of-Wight-sized lump of DU, anyway? Ben
Re: Stokers Required
On Mon, Feb 24, 2003 at 04:37:03PM +, Robin Szemeti wrote: > > I think the technical term for this is "JFDI" ... :) ... anyway .. I say do > it. Its something we've talked about a lot in the past both on here and in > 'another place' .. there is mileage in it ... let the party commence. I'm starting to think that JFDI, just like SWS and Interesting Times, is actually a virus. Someone (looking at no-one in particular) seems to have infected me with JFDI recently. Ben
Re: Anyone have a spare Sun Keyboard ?
On Mon, Mar 10, 2003 at 12:55:24PM +, Lusercop wrote: > On Mon, Mar 10, 2003 at 12:24:59PM +, Leo Lapworth wrote: > > Yes, it has been said I should use serial console > > instead but as I'm doing a complete reinstall > > I'd rather use a keyboard. > > This is a non-sequitur. Please explain. Actually, I'd agree with Leo. I've seen Solaris installs over console go horribly evilly wrong (Yes, I know that they *shouldn't* but) Ben
Re: Anyone have a spare Sun Keyboard ?
On Mon, Mar 10, 2003 at 01:20:54PM +, Lusercop wrote: > On Mon, Mar 10, 2003 at 01:22:58PM +0000, Ben wrote: > > On Mon, Mar 10, 2003 at 12:55:24PM +, Lusercop wrote: > > > On Mon, Mar 10, 2003 at 12:24:59PM +, Leo Lapworth wrote: > > > > Yes, it has been said I should use serial console > > > > instead but as I'm doing a complete reinstall > > > > I'd rather use a keyboard. > > > > > > This is a non-sequitur. Please explain. > > > > Actually, I'd agree with Leo. I've seen Solaris installs over console go > > horribly evilly wrong (Yes, I know that they *shouldn't* but) > > OK, I'm surprised at this. I had no problem doing a solaris install like > that, though I may try and do a jumpstart for the one that I'm about to > do. Errm, you may want to re-read what I wrote, ducks. I have seen Solaris installs over console go wrong. That does not imply that every install I've done like that (this is over a data set of at most four points) has gone wrong. I just prefer not to chance it. Ben
Re: Anyone have a spare Sun Keyboard ?
On Mon, Mar 10, 2003 at 02:01:43PM +, Bob Walker wrote: > On Mon, 10 Mar 2003, Ben wrote: > > > > I have seen Solaris installs over console go wrong. That does not imply > > that every install I've done like that (this is over a data set of at most > > four points) has gone wrong. I just prefer not to chance it. > > this is not the consoles fault or indeeed the solaris installers fault. I dispute that it is not the fault of the installer. I am not careless, incompetent or inexperienced with Solaris. If the installer is still tripping me up, I submit it may be misdesigned. > this is user error. That is not in doubt. Personally, when I'm doing a procedure which has a possibility of going wrong due to my error, I may want to find a method which I am more likely to execute correctly. Why would I want to take risks which are neither necessary nor fun? Ben
Re: Anyone have a spare Sun Keyboard ?
On Mon, Mar 10, 2003 at 06:30:18PM +, Steve Mynott wrote: > > On Monday, Mar 10, 2003, at 14:15 Europe/London, Ben wrote: > > > I dispute that it is not the fault of the installer. I am not careless, > > incompetent or inexperienced with Solaris. If the installer is still > > tripping me up, I submit it may be misdesigned. > > What was the actual problem you saw? It was about 2.5 years ago, so I can't really remember. I'd started the install because my SAs were flat out, and found myself with a machine that would not boot, and which no longer recognised the console. I handed it over to one of my team who IIRC had to get a working Solaris install off another machine, put the HD into the naughty one, get hold of a keyboard from somewhere, and then sort out the b0rked drive. This added the best part of a day to the install and nearly caused us serious problems with the timeline. After having been bitten like that, I err on the side of caution. I accept that it was exceptional circumstances, but I prefer to be cautious where production kit is involved. It also isn't the only operating system I've seen have that kind of exceptional craziness. I have seen an upgrade of SuSE to Debian produce a machine that will boot, but claims to have 255 imaginary SCSI devices (and probes each one before continuing) and entirely fried PCI data in /proc. That was ... unpleasant. > BTW I have a Sun keyboard (Type 5c) you can borrow if you can collect > from Stoke Newington. Thanks, but it was Leo that wanted it, not me. Ben
Re: Obsolete software
On Fri, Mar 14, 2003 at 05:28:18PM +, Chris Benson wrote:
> On Fri, Mar 14, 2003 at 02:44:07PM +, Frank Booth wrote:
> >
> > An Oracle query will perform with varying degrees of efficiency,
> > recognising useful indexes etc, based entirely on the sequence of the
>
> Bzzzt! in 8i and 9i, Oracle defaults to the Cost Based Optimizer.
> Unfortunately it needs a (manual)
>
> SQL> exec dbms_utilities.gather_schema_statistics('APPS');
>
> (or something like that) to work reasonably well. Viz last week a
> 4-table query on out of date statistics -> 15min to use 500MB of temp
> space then fail. With current statistics: -> correct results in 0.08
> seconds :-}
The CBO in Oracle 8.1.6 is also still on a fair amount of crack. I've seen
it just totally fail to utilise an index. Rule hint is sometimes the
only way to stop it behaving like a knobgoblin.
Happily, Oracle have decided to drop the RBO from the next release of
Oracle. IIRC it's already marked as 'deprecated'. Well, make the CBO
work properly then, motherfucker!
Ben, having been wrestling with the optimizer for a while now.
Re: crackfuelled idea / nntp / message boards / bl*gs / mailing lists
On Fri, Mar 28, 2003 at 12:18:27AM +, Simon Wistow wrote: > > So it would be a mail <-> blog gateway. > > Put in a mail <-> news gateway as well and suddenly, as I said somewhere > else, your stading on the platform at the station of Convergence City > with a ticket to Cracktown and Buzzwordville. And then you realise that you're going to be sharing a carriage with fifty Coates-Hammersley groupies. You wake, bathed in sweat, the only sound the pounding of your own fevered heart. Your adventure is over. Would you like to try again? Only this time, try and Choose Your *Own* Adventure. Ben
Re: The joys of web development
On Wed, Apr 02, 2003 at 07:49:27AM -0800, jonah wrote: > > Website accessability to the disabled is one of my little pet rants, but I > won't bore you lot with it. You're all much better at ranting than I am. *waves sushi around until Marna notices* Kitty
Re: international beer summit
On Wed, Jun 04, 2003 at 11:43:43PM -0400, Perrin Harkins wrote: > On Wed, 2003-06-04 at 19:47, Kate L Pugh wrote: > > Yes! > > Great! I'll be there from the 22nd through the 28th and have no plans > yet. I'm going to rent an apartment with some friends in Clerkenwell. As Glastonbury (to which a fair few l.pm'ers are going) is the 27-29th of June, and many people travel up on the 26th, might I suggest either the 25th or 24th as possible dates? Ben
Re: The Perl Color?
On Thu, Jun 05, 2003 at 02:40:20PM +0100, Ali Young wrote: > > OK, lets solve this: > Take a large quantity of psychotropic drugs. > Look at some Perl code. Curse you! I'm working this weekend Ben
Re: Dim sum today
On Wed, May 28, 2003 at 12:13:11PM +0100, Earle Martin wrote: > On Wed, May 28, 2003 at 11:52:47AM +0100, Nicholas Clark wrote: > > Grubstreet has no entry for New World :-( > > It does now. > > http://grault.net/cgi-bin/grubstreet.pl?New_World,_W1D_5PA I'd like to contribute a few pages. Can someone give me a headsup on how to make them so that they fit into 'House Style'? Ben
weird eval
What circumstances are there under which eval {}; will not trap a program exit ?
I assume a naughty XS module segfaulting will do for it - but are there any
others?
Ben
Re: MS Office, Klingon Edition
On Sat, Jun 07, 2003 at 10:43:59PM -0400, R. Geoffrey Avery wrote: > At 05:33 AM 6/6/2003, you wrote: > >Andrew Beattie wrote: > >>Philip Newton wrote: > >> > >>> Microsoft's Office suite currently ships in 18 languages > >>MS-Office would be more fun in Klingon. > > But if you are a Klingon with mental health problems, induced by Microsoft > or otherwise, you should go to Oregon where they will have an interpreter > at (the) hospital for you. This hasn't made it on to snopes yet, I think, but there's an excellent debunking of this latest urban myth on David Farbers IP (www.interesting-people.org) list. Ben
Re: Templating systems techtalk
On Fri, Apr 18, 2003 at 02:00:09PM -0400, darren chamberlain wrote: > If you think C++ is not overly complicated, just what is a protected > abstract virtual base pure virtual private destructor, and when was > the last time you needed one? > -- Tom Cargill That is mis-attributed. It's actually by Peter van der Linded, from his book 'Expert C Programming', which I highly recommend. Ben
[REVIEW] Graphics Programming With Perl
Found this mostly finished on my HD. Finished it off and thought I'd post it - I think someone may have already posted a review of this, though. Ben =head1 NAME Graphics Programming With Perl =head1 Author Martien Verbruggen =head1 ISBN 1-930110-02-2 =head1 REVIEW On the face of it, the subject matter leads a lot to be desired. Perl is not the best choice of language for the sorts of operations that graphics seem to require. As the author notes, however ,the ability to leverage meaty XS extensions allows Perl to gain much of the low-level power and speed of C. This enables the Perl module-writer to compensate for the overly-meta way in which Perl often deals with binary data. For the most part, Verbruggen deals with his subject matter very well. The opening chapters cover basic graphics concepts, colour spaces, formats, etc and help set the tone for the rest. It is clear that this book is intended as a jumping-off point and a cookbook, rather than a reference guide or authorative tome.This is something of a shame as Verbruggen's style is open and easy to read, so a reference section written by him would likely have been readable and usable. In addition, much of the focus is around web-based graphics but that surely just reflects the needs of its target audience. Virtually all of the second part of the book is taken up with discussing the sort of effects and graphics creation a web programmer might want to achieve. In practice, on reading it, I found quite a bit of the coverage to be somewhat superficial. The major models have their simple operations well-covered, but there was little of the special knowledge or detail of the quirks and secrets that can really mark a tech book out as very special. On the upside, the issues of version numbering and nasty dependency hell (which have often beset my oiwn experiences with Perl and graphics) are very well-handled and would doubtless save a less-experienced programmer a lot of heartache. The occasional shallowness of the material is my only real concern about this book. In particular, I found the discussion of the Gimp rather disappointing and that the author rather dropped the ball on SDL. The Simple Direct media Layer has opened the possibility of entire graphics-heavy applications being written in Perl, and would have been an ideal candidate for a chapter. In addition, some discussion as to why an application designer might want to employ the direct methods of OpenGL while within a windowing environment would have been interesting, given the poor coverage it has received in several other recent books. One other nit was the authors somewhat peculiar dialect of Perl. Possibly this should have been caught at the tech reviewing stage, but very little of the example Perl in the book will run on a Perl earlier than 5.6.0 - but in most cases this is just because the author uses 'our' liberally. This directly contradicts the discussion in the introduction about Perl versions, which is very bad. The dialect has other peculiarities as well - bare filehandles instead of IO::* modules, extensive use of the CGI module - despite its known systemic problems with memory leaks, insistence on methods being available as class or instance methods wherever possible, but then introducing possibly troubling calling semantics by passing @_ unmodified by delegation, use of @ISA instead of 'use base', using the -w switch instead of scoping warnings, etc. The book also contains some factual errors - for example Macromedia do *not* require a license fee for software which produces SWF files unless it contains derived code from their reference implementation. Like some of the other works in the Manning stable, this book suffers at times from being neither fish nor fowl - failing to provide enough reference material for longevity, but sometimes requiring too much background. Hopefully, a second edition would allow more space for the reference section this book deserves. Despite this, in many ways Verbruggen provides a charming introduction to the topic of graphics and Perl. He covers some interesting modules which may have been unknown to the reader and the book should be an excellent read for the intermediate programmer with an interest in graphics.
501 Not Implemented
Hi, I'm having a nasty HTTP implementation mismatch. Some server I'm talking to using LWP (I can't tell what the server actually is, and Netcraft don't know either) is returning a 501 to a POST, not sending any response body and then closing the connection. (As a side effect, this is causing a segfault somewhere down in the XS code, but I think I understand how to workaround this and will isolate the cause and try and patch it later). Now, I know that a 501 SHOULD contain a response body, but that's kind-of not relevant. What I want to know is what server conditions could cause it to think that a 501 is an appropriate thing to send back. I know that Not Implemented could apply to a request method or a transfer-coding but are there any other examples that people know of that could trigger this? My current theory is that the server can't handle the transfer-coding and is responding normally, if in a somewhat clipped dialect. I don't believe that it can't handle POST as a request method. (I think that the segfault is being caused by fishyness in the way the other side is doing TLS, but that's another story). Thoughts? Am I on crack this morning? Ben
Re: 501 Not Implemented
On Mon, Jun 16, 2003 at 07:51:54PM +0100, Patrick Mulvany wrote: > On Mon, Jun 16, 2003 at 11:32:13AM +0100, Ben wrote: > > Hi, > > > > I'm having a nasty HTTP implementation mismatch. > > > > Some server I'm talking to using LWP (I can't tell what the server actually is, > > and Netcraft don't know either) is returning a 501 to a POST, not sending any > > response body and then closing the connection. (As a side effect, this is causing > > a segfault somewhere down in the XS code, but I think I understand how to > > workaround this and will isolate the cause and try and patch it later). > > > > Now, I know that a 501 SHOULD contain a response body, but that's kind-of not > > relevant. What I want to know is what server conditions could cause it to > > think that a 501 is an appropriate thing to send back. > > > > I know that Not Implemented could apply to a request method or a transfer-coding > > but are there any other examples that people know of that could trigger this? > > Are you setting the Referer correctly for those pages? Ah, you see, this isn't a page which a browser would ever see. It's an application that I POST some XML to, and it gives me back another XML document based on what I send it. I have a document which describes how this thing is supposed to work, and it doesn't behave the way it's supposed to. If I send it garbage, well-formed but invalid XML or correctly formed but incorrect XML, then I get an error message. If I send it something correct, I get a 501 and no return document. Which is why I think it shouldn't be the transfer-coding or anything else. It clearly *can* interpret my requests, as it knows what incorrect ones look like. > Best bet it to browse it while capturing the datastream. > Then you really know what it will accept rather that what you > thought it needed. Then you can backout to a minimum data request. I've tried that. Same behaviour in the browser. I'm now pretty sure the problem is with them, but the only contacts I have are non-technical. I tried asking for someone technical's email address and got told that my existing contact would be happy to answer my questions, which *really* isn't what I wanted. Most annoying, all round. Ben
Re: SQL standards
On Sat, Jun 21, 2003 at 03:45:29PM +0100, Shevek wrote: > > I'm doing a fairly hardcore security analysis so I'm trying to find > canonical references for the security systems, with full descriptions of > the semantics in the various servers. So in the case of SQL92/99, I need > the precise wording of the standard. The standards, especially SQL 99 are just horrible. Because of the mutually entirely different standpoints taken by the major vendors, the standardising body just seemed to opt for a standard which represents the union of all major implementations. Of course, to then be able to allow any vendor to claim 'compliance' with the standard you then have to redefine compliance to mean "implements one of a number of different possible semantics for quite core elements". Good Luck. I expect you'll need it. If you want to talk more about the little bit of this that I know about, mail me offlist. Ben
Re: pub recommendation near Northampton sq
On Mon, Jun 23, 2003 at 10:38:49AM +0100, Roger Burton West wrote: > On Mon, Jun 23, 2003 at 09:25:33AM +, Martin Bower wrote: > >anyone know any decent pubs within walking distance of City University ? > >http://www.ukuug.org/events/TimOReilly/ > > Only the Wenlock, and if there are more than about four of you it's not > really suitable. > > http://grault.net/cgi-bin/grubstreet.pl?Wenlock_Arms,_N1_7TA I like the Old Red Lion, which is here: http://www.streetmap.co.uk/newmap.srf?x=531475&y=183099&z=1&sv=531250,183250&st=4&ar=Y&mapp=newmap.srf&searchp=newsearch.srf Upper St is also not too much further than the Red Lion, although I find many of the pubs up there to be Not Great. Ben
Re: [ot] Mounting Unix Drives in Windows
On Tue, Jul 01, 2003 at 11:49:02AM +0200, Rafael Garcia-Suarez wrote: > Dave Cross wrote: > > So, a few questions: > > > > 3/ Are there any other solutions we can look at - like, perhaps, > > an NFS client for Windows? > > There are NFS clients for Windows. Alternatives : > WebDAV ? IIRC Windows' file explorer supports it natively. WebDAV, especially M$'s implementation of it, should *not* be used for this sort of file transfer. Especially not for large files. filesystem-over-HTTP is not necessarily as good an idea as it might first appear. As others have noted, Samba really is the solution here. Ben
Re: [ot] Mounting Unix Drives in Windows
On Tue, Jul 01, 2003 at 12:09:56PM +0200, Rafael Garcia-Suarez wrote: > Robin Berjon wrote: > > Rafael Garcia-Suarez wrote: > > > Alternatives : > > > WebDAV ? IIRC Windows' file explorer supports it natively. > > > > It does. I'm not sure however that using WebDAV as a full time file system is a > > happy choice, not sure it's been tested that intensively. It'll also be slow. > > So WebDAV, like Samba, doesn't require any additionnal software on the > recent clients. Good point for it. That depends on the quality of the client which comes pre-installed. > The relative performances depends on the use case. I believe that WebDAV > demands less bandwidth than Samba. That depends very sensitively on the combination of client and server, and how they interpret the semantics of the incomplete spec that is WebDAV. For example, under some circumstances, some versions of the M$ DAV client follow a DAV-enhanced PUT /foo with a GET /foo to verify that the PUT succeeded. This doubles the user-perceived time taken to complete a file upload. In general, I'm unhappy with a solution which has this sort of nasty mismatch of implementations possibility.[0] > I'm quite confident in the quality of mod_dav (for apache 2) due to my > experience with Subversion. But that's a personal impression. We don't > use WebDAV by itself at work -- just as a support layer for DeltaV, the > protocol extension used by Subversion over http. Subversion and DeltaV really aren't that happy a marriage, or weren't the last time I looked at them. The problem is that completely general version control systems are actually remarkably difficult to get right, due to the entirely different (and at times impossible to reconcile) usage modes that people want to use them in.[1] Subversion and DeltaV have different ideas about some fairly fundamental things, IIRC. > Now there's also the questions of access control, file ownership, rights > management, etc... which are (to my taste) easier to manage in Apache > than in Samba. YMMV. Ben [0] OK, OK. So I should really stop using SSL, then... [1] If anyone's tempted to follow up with "I don't see what's so difficult about general VC systems...", I'd advise them to go read the IETF DELTAV WG archives. See you in a while.
Re: Linux firewall / web server
On Mon, Jun 30, 2003 at 07:38:12PM +0100, Chris Benson wrote: > > I had a lot of difficulty thinking about the f/wall rules for a system > acting as f/wall and server until I separated the data streams (and > setup a table/chain for each stream): > > inet-> f/wall > inet-> internal network > inet-> dmz network > dmz -> inet > dmz -> internal > dmz -> f/wall > int'l -> f/wall > int'l -> dmz > int'l -> inet > f/wall -> inet > f/wall -> dmz > f/wall -> internal > > I've a perl script that writes a set of iptables commands from a > simplified config file ... People might also be interested in filtergen (http://hairy.beasts.org/filter/ also comes as a Debian package and probably an RPM). It was written by a mate of mine, it handles more than just iptables as a backend and I'm reasonably happy with it. Ben
Re: Linux firewall / web server
On Tue, Jul 01, 2003 at 12:47:40PM +0100, Shevek wrote: > On Tue, 1 Jul 2003, Ben wrote: > > > On Mon, Jun 30, 2003 at 07:38:12PM +0100, Chris Benson wrote: > > > > > > I had a lot of difficulty thinking about the f/wall rules for a system > > > acting as f/wall and server until I separated the data streams (and > > > setup a table/chain for each stream): > > > > > > inet-> f/wall > [YADDA] > > > f/wall -> internal > > > > > > I've a perl script that writes a set of iptables commands from a > > > simplified config file ... > > > > People might also be interested in filtergen (http://hairy.beasts.org/filter/ > > also comes as a Debian package and probably an RPM). It was written by > > a mate of mine, it handles more than just iptables as a backend and I'm > > reasonably happy with it. > > I do not understand the need for [the added complexity and perversion of] > such packages since it is perfectly possible to write something almost > syntactically identical in the shell anyway using a few shell variables. I disagree. The syntax of something like filtergen is obvious at a glance, and greatly simplifies the audit process. The filtergen rules can be audited and seen to be correct by someone who is a great deal less competent and knowledgable than the person who wrote them (but who has to support them). I submit that the case where a system is maintained and supported by people less gifted / experienced than those who architected and built it is sufficiently common that tools that make this easier are useful. But, in any case, I'd be interested in seeing the shell stuff that you'd use to accomplish this task. Ben
Re: Linux firewall / web server
On Tue, Jul 01, 2003 at 01:21:03PM +0100, Shevek wrote: > On Tue, 1 Jul 2003, Ben wrote: > > > On Tue, Jul 01, 2003 at 12:47:40PM +0100, Shevek wrote: > > > I do not understand the need for [the added complexity and perversion > > > of] [firewall rule management] packages since it is perfectly possible > > > to write something almost syntactically identical in the shell anyway > > > using a few shell variables. > > > > I disagree. The syntax of something like filtergen is obvious at a glance, > > and greatly simplifies the audit process. The filtergen rules can be audited > > and seen to be correct by someone who is a great deal less competent and > > knowledgable than the person who wrote them (but who has to support them). > > > > I submit that the case where a system is maintained and supported by people > > less gifted / experienced than those who architected and built it is > > sufficiently common that tools that make this easier are useful. > > > > But, in any case, I'd be interested in seeing the shell stuff that you'd use > > to accomplish this task. > > The second point does rather destroy the validity of the first! Hardly. I'm always interested in alternatives to my current way of doing things to see if improvements can be made, or if I've missed out on interesting things. > Anyway, even having looked at filtergen, who really gives a toss if it > came in on eth0? That's assembly programming for firewalls. What everyone > really wants to say is, "If it is going to our company web server, let it > in." and "eth0" is so far separated from any such concept that I disagree > that "eth0" should EVER appear in a firewall, regardless of how clever the > syntax. Nonsense. I am *very* interested in knowing which interface a packet that *claims* to come from my internal network was received on. [snip code example] Well, yes, you can do that, but for a rulebase of the complexity on the order of what I want, this is going to degenerate into an unmaintainable quagmire. In fact, when one of my SAs decided it would be good to try and remove filtergen and "just have an rc.firewall script instead" it *was* an unmaintainable quagmire. Ben
Re: Linux firewall / web server
> > > Anyway, even having looked at filtergen, who really gives a toss if it > > > came in on eth0? That's assembly programming for firewalls. What everyone > > > really wants to say is, "If it is going to our company web server, let it > > > in." and "eth0" is so far separated from any such concept that I disagree > > > that "eth0" should EVER appear in a firewall, regardless of how clever the > > > syntax. > > > > Nonsense. I am *very* interested in knowing which interface a packet that > > *claims* to come from my internal network was received on. > > Note that I stated that there was a validation chain. Yes, of course you did. I missed that on first reading, sorry. Post-Glastonbury and all that *waves hands*. I'd still rather have all the rules in one place, than separate chains and so on. People have been known to become confused and have conceptual problems with chains, etc. This may say more about the people I have administering firewalls than anything, though. Ben
Re: Detecting + Preventing RAM bloat
On Thu, Jul 03, 2003 at 07:56:31AM -0500, Nigel Hamilton wrote: > > Any good defensive programming practices that help to avoid > writing memory-hungry programs? For example, at the moment I'm just > letting object references go out of scope, should I be using DESTROY? One which springs to my mind is that as arrays and hashes are never shrunk, then you may wish to add levels of indirection so that large-but-volatile arrays and hashes are one level of indirection deeper than the rest of the data structure, so that when the reference count of the deeper hash / array goes to zero, the dictionary space used by the large-but-volatile stuff can be reclaimed. Depending on the relative sizes of the dictionary and the average key and value length, this may or may not be a sensible idea, however. No warranty, YMMV, blah blah #include , etc Ben
Re: parsing lisp s-expressions in perl?
On Wed, Jul 16, 2003 at 03:04:07PM +0200, Raf wrote: > > Just a quick query to find out if there's a good tried and tested module > for parsing lisp s-expressions into perl datastructures (or eventually > into xml?)? Don't know how complete it is, but Paul Crowley's unreleased SPKI::Sexp might do what you want. I'll just check with him that it's OK to distribute. Ben
[OT] nasty mod_perl build problem
Hi, Anyone know anything about this load error from mod_perl: Cannot load /usr/local/apache/libexec/libperl.so into server: /usr/local/apache/libexec/libperl.so: undefined symbol: Perl_Ipatchlevel_ptr I'm running apache 1.3.28 and mod_perl 1.28 and the build seems to happen fine, but it just won't load the .so I can't find the symbol anywhere obvious at all. Meep. Any ideas? Ben
Re: [OT] nasty mod_perl build problem
On Fri, Jul 25, 2003 at 10:08:27AM +0100, Andy Wardley wrote: > Ben wrote: > > Cannot load /usr/local/apache/libexec/libperl.so into server: > > /usr/local/apache/libexec/libperl.so: undefined symbol: Perl_Ipatchlevel_ptr > > Could it be that you're using an Apache/mod_perl that was built with one > version of Perl, and you've later installed a new version of Perl? The perl is the current Debian unstable (5.8) and the Apache and mod_perl were built by myself[1] from source. *grumble* Ben [1] Don't say "Why don't you just use the debian-supplied ones" to me unless you actually want to hear the answer. Yes, apt / dpkg gets a lot of things right. Two of the things it doesn't get right IMNSHO are Apache and Perl.
Re: [OT] nasty mod_perl build problem
On Fri, Jul 25, 2003 at 01:26:26PM +0100, Richard Clamp wrote: > On Fri, Jul 25, 2003 at 11:24:47AM +0100, Ben wrote: > > [1] Don't say "Why don't you just use the debian-supplied ones" to me unless you > > actually want > > to hear the answer. Yes, apt / dpkg gets a lot of things right. Two of the things > > it doesn't > > get right IMNSHO are Apache and Perl. > > I don't think it gets perl right at all, or at least not the way I use > it. That's why I typically install into /usr/local/perl*[0] and if I > need it, I build mod_perl against the debian apache as a dso. Plan. I think I'll do that. Richard C in "being sensible" shocker. Ben
Re: [OT] nasty mod_perl build problem
On Fri, Jul 25, 2003 at 01:43:30PM +0100, Ben wrote: > On Fri, Jul 25, 2003 at 01:26:26PM +0100, Richard Clamp wrote: > > On Fri, Jul 25, 2003 at 11:24:47AM +0100, Ben wrote: > > > [1] Don't say "Why don't you just use the debian-supplied ones" to me unless you > > > actually want > > > to hear the answer. Yes, apt / dpkg gets a lot of things right. Two of the > > > things it doesn't > > > get right IMNSHO are Apache and Perl. > > > > I don't think it gets perl right at all, or at least not the way I use > > it. That's why I typically install into /usr/local/perl*[0] and if I > > need it, I build mod_perl against the debian apache as a dso. > > Plan. I think I'll do that. ... and it worked like a charm. *pint*. Cheers, dude. Ben
Re: Siesta party
On Thu, Aug 14, 2003 at 10:54:03AM +0100, Nigel Rantor wrote: > > Sorry, I'm a bit of a foodie. (AAnd if you like steak, then you should > go to the camden brasserie, but thats a whole NOTHER story...) Are you on london.food ? Ben
Re: [JOB] Java dev @ Kizoom
On Tue, Aug 12, 2003 at 12:43:33PM +0100, Steve Purkis wrote: > Not exactly a Perl role, but then I suppose that makes this more or > less on topic for this list ;-). > > Kizoom are looking for a Java developer with a minimum of 3 years > recent experience to join their team to do s/w development for mobile > devices. Solaris admin experience is a definite plus. You can read > more about the company at: > > http://www.kizoom.com > > If you're interested (or know somebody who is), send your CV to: > > Adam Cohen-Rose <[EMAIL PROTECTED]> I've never had any contact with them personally but several friends and ex-colleagues work, or have worked there, and they seem to be reasonable people. Ben
Re: Oops - I meant OT Virtual Reality?
On Wed, Aug 06, 2003 at 04:29:42PM +0100, Paul Golds wrote: > > Naturally I skip over such posts anyway, nasty little things. Hopping around, eating carrots. Ben
Re: text'd or texted
On Tue, Aug 05, 2003 at 09:48:59AM +, Dominic Mitchell wrote: > On Tue, 5 Aug 2003 04:02:23 -0500, Elaine -HFB- Ashton <[EMAIL PROTECTED]> wrote: > > Welcome to English where any noun can be verbed :) Also popular is 'to > > SMS' or 'to be SMS-ed'. > > Surely that's just a smidgeon to close to S&M'd? (Assuming s/ to / too / ... ) I've never seen that particular verbing before. The verbs 'do' or 'play' are generally far more likely to be used in that context. YMMV, of course, but just because any noun can be verbed doesn't make it sensible to do so. Ben
Re: OSX - 'the real question'
On Wed, Aug 20, 2003 at 11:17:14AM +0100, Sam Vilain wrote:
>
> The consensus seems to be to *always* leave the system software
> (Perl et al) alone, and put your stuff in a custom directory such
> as /usr/local, /opt, or /sw. Tinkering with the contents of
> /{usr,}/{bin,sbin} is generally seen as something that shouldn't
> be done in most cases.
>
> I think a better way to put this is that all changes to the OS owned
> directories should be made through the OS' native packaging system.
That rather depends upon where one chooses to draw the dividing line
between the OS and the applications which live on top of it.
Ben
Re: XML & XML::LibXML declarations issue
On Wed, Aug 27, 2003 at 12:52:48PM +0200, Robin Berjon wrote: > Nigel Hamilton wrote: > > For example, a bank I previously worked at has trashed a lot of its SOAP > > transaction system because it was too slow. The reason? The byte size of > > the XML wrappers dwarfed the actual payload of data. > > > > Too much time was being spent stuffing the XML 'envelope', posting it, and > > opening it up on the other side. > > Yes, I'm aware of the issue, which is probably why I ended up on the program > committee of this[0] Thing With A Horrible Name :) > > [0]http://www.w3.org/2003/07/binary-xml-cfp.html Pardon me for being thick here, but what possible gains are there for this in general-purpose use over just gzip / zlib ing the data ? IME the parsing of a non-trivial binary format is a much nastier task than parsing text, especially structured text like XML. Vendors are generally pretty shit at providing even a decent interoperable *text*-based protocol (viz variously and gloriously broken HTTP/1.1 implementations (c) All Major Vendors). HTTP is, I would submit, hardly a particularly difficult or fragile protocol to get right, yet vast numbers of software companies seem incapable of getting it right or handling its edge cases sanely. On the binary format front, would we like to consider the resounding triumph of clarity and interoperability represented by CORBA or RMI-IIOP or any of the other nasties out there? I can't see that reinventing the wheel with a bunch of domain-specific binary formats is anything other than a retrograde step. What am I missing? Ben
Re: XML & XML::LibXML declarations issue
On Wed, Aug 27, 2003 at 02:16:32PM +0200, Robin Berjon wrote: > Ben wrote: > > On Wed, Aug 27, 2003 at 12:52:48PM +0200, Robin Berjon wrote: > >>[0]http://www.w3.org/2003/07/binary-xml-cfp.html > > > > Pardon me for being thick here, but what possible gains are there for this > > in general-purpose use over just gzip / zlib ing the data ? > > I tend to put speed, size, random access, dynamic update on my list of benefits. > Other people have other views (often a subset of that). Sometimes gzip doesn't > cut it, either because even with higher CPU you may have battery-usage problems > (eg on cell-phones) or because you want much better compression (I've gotten > x200 compression on some SOAP messages where gzip got around x3 for intance). > > Random access has all sorts of uses, for instance it's dead useful for an XML > editor (it can change a document in the middle of it, instead of reading it all, > changing it, and writing it all out). Dynamic update is very good for very low > bandwidth situations (where very low can be what you'd call really low, or can > be something much bigger but you have lots of data to send). It is also critical > for unidirectional systems such as digital TV broadcasts (putting SVG on the telly). This would be essentially a form of chunked parsed encoding, no? The idea being that an app continues to chuck around its data structures in (some possibly trivially transformed version of) the wire format? So that an 'XML editor' really does mean that (in some sense) the data is represented as XML all the way down into the Model layer. Is that the idea with dynamic update? > > I can't see that reinventing the wheel with a bunch of domain-specific binary > > formats is anything other than a retrograde step. > > We're not gunning for domain-specific. We're gunning for generic. It won't > happen if it's not simple. http://www.w3.org/2003/07/binary-xml-cfp.html says: "3. Improved network efficiency and reduced storage needs: compression techniques that make use of domain-specific knowledge often do better than more generic compression;" Robin says: "I've gotten x200 compression on some SOAP messages where gzip got around x3 for intance" I said: "in general-purpose use" I'll be interested to see the results, but as gzip is, basically optimal *in the general case* I don't think that a generic scheme which doesn't degenerate into a bunch of special cases will work properly. I don't doubt for a minute that certain types of specialised format / compression are a good idea for certain specific applications and domains - it's the genericity I question. Coming back to SWF - this really should serve as a warning to anyone planning to design a binary file format or protocol. I'm almost of the opinion that you shouldn't be allowed to do so without first having implemented a parser for that abortion[0]. The most important salient point is: * No matter how much they cry, no matter how much they beg - never, ever optimise for a low bandwidth link without a sane extension framework and a way to be forward compatible for the day when low bandwidth has ceased to be an issue. > PS: Ben, I've answered your other private mail but your SMTP server blacklists > the entirety of wanadoo.fr, which honestly is quite a bad idea :) I'll have a word with the person who runs the box. Thanks. Ben [0] http://www.openswf.org/ for the trail of tears...
Re: golf and reversed emails
On Thu, Aug 28, 2003 at 10:04:43AM +0200, Philip Newton wrote: > > Reminds me of some programming contest where you had to take as little > time as possible to do task X. You were allowed to submit multiple > candidates; the one that took the least amount of CPU time would > determine your score. > > I believe the winner was one that submitted n candidate programs, n > > 2; one of them actually completed the task and wrote the correct result > to a file, another simply slept, occasionally looking to see whether > this special file existed. When it showed up, it would print the > result. Bingo: correct result with very little CPU time. > > (There was another chappy who munged around in memory, overwriting some > OS bookkeeping values, which made it look as if his program took > negative CPU time.) > > They tightened up the rules for subsequent contest, I believe... Yes. I first read about this is in Peter van der Linden's excellent book Expert C Programming. I would heartily recommend it as one of the most readable books about a programming language I have ever read. Go find. Ben
Re: Fave calendering software?
On Fri, Aug 29, 2003 at 12:11:05AM +0100, Paul Makepeace wrote: > Basic hour-by-hour, daily, weekly, monthly, yearly views. Something that > produces HTML output for inclusion or direct embeddable on the web would > be my personal ideal, to be shared with various types of people. E.g., a > client could see in detail what I'm doing on their project and the rest > is simply blocked off as "available/not available". Prospects could get > an overview of availability say during a month. There's always swarmcal (written by Simon Wistow and my evil twin, Skippy). It's *so* not finished but it's up and working - in fact quite a few l.pmers have accounts on it. Mail me offlist if you want details / code. Ben
Re: London.pm identity cards
On Fri, Aug 29, 2003 at 10:11:13AM +0100, Sam Vilain wrote: > > This year's modesty award is given for a phrase spoken by a lecturer > after a rather difficult concept had just been introduced. > " You may feel that this is a little unclear but in fact I am > lecturing it extremely well." That sounds like a quote from TWK, or someone like him. That's the second pseudo-Kornerism you've sigged us with today. Is there a reason? Ben, curious now
Re: gzipping your websites
On Mon, Sep 01, 2003 at 12:40:53PM +0100, Simon Wistow wrote: > > Now, gzipping your outgoing webpages is a good thing cos it cuts down on > transmission time and reduces your bandwidth costs. > > Unfortunately it's also fairly processor intensive. > > So I'm looking for a external solution and wondered if people had had > any experience. > > The solutions I've looked at so far are I evaluated the same problem set a while ago. If you have the budget, I would seriously consider the RedLines. They *are* expensive, but they do more than just external compression. They do load-balancing as well, and in a number of fairly typical situations remove the need for a reverse proxy layer. As always, it depends what you're trying to accomplish. The group of sites I was working on were hosted in an environment where bandwidth is expensive. The compression (and reporting stats so we knew which sites and which parts we needed to optimise) plus the removal of the proxy layer (and the removal of our previous load-balancing tech, which had been EOL'd) made the cost actuallly worth it. Mail me offlist for more details, especially if you're seriously considering Redline. Ben
Re: Mercury Amalgam (was: insidious biometrics, identity crises)
On Tue, Sep 02, 2003 at 03:48:06PM +0100, Philip Hellyer wrote: > > My s.o. also took 10 grams of intravenous vitamin C during > the procedure. Isn't that above the level where vitamin C will crystalise out of urine and potentially cause damage to the urethra? I mean, the RDA for vitamin C is only, what, 60 milligrams? This seems awfully high to me. Ben
Re: Bad C Source (Re: gzipping your websites WINRAR 40 days trial)
On Thu, Sep 04, 2003 at 01:03:16PM +0100, Lusercop wrote:
> On Wed, Sep 03, 2003 at 11:48:19AM +0100, Simon Wistow wrote:
> > I have to admit, I like gotos in C. This is not a winning testimonial
> > though. I've been told that my C is like Object Orientated assembler
> > which is fair enough because I learnt C after I'd learnt 68k.
>
> Hmmm, I like gotos too, but mainly for freeing.
>
> [deliberately not snipped for context]
> > My code typically looks something like ...
> >
> > quirka_t *
> > foo (foo_context * context, int * error)
> > {
> > /* malloc bar and if that fails set error and goto FAIL */
> > /* malloc quirka and if that fails set error and goto FAIL1 */
> > /* do something with quirka and if that fails, set error, goto FAIL2 */
> >
> > SUCESS:
> > return quirka;
> >
> > FAIL2:
> > /* free quirka */
> > FAIL1:
> > /* free bar */
> > FAIL0:
> > return NULL;
> > }
> > apparently people ahve been known to be sick.
>
> That is pretty ugly. The thing to remember is that in ANSI, "free(NULL);" is
> a) valid, and b) a NOOP. So I'd write that as:
>
> | quirka_t *foo (foo_context *context, int *error) {
> | bar_t *bar=NULL;
> | quirka_t *quirka=NULL;
> |
> | bar=(bar_t *)malloc(sizeof(bar_t));
> | if(!bar) goto error;
> |
> | quirka=(quirka_t *)malloc(sizeof(quirka_t));
> | if(!quirka) goto error;
> |
> | /* do stuff */
> |
> | end:
> | free(bar);
> | return(quirka);
> | error:
> | free(bar);
> | free(quirka);
> | return(NULL);
> | }
This has the problem that if you're calloc'ing and initialising a complex
structure then the free's aren't necessarily that simple. Eg:
miner_t *smith (surbiton_context * context, int * error) {
miner_t * foo = NULL;
if ((foo = (miner_t *) calloc(1, sizeof(miner_t))) == NULL) {
*error = ENDORIAN;
goto FAIL1;
}
if ((foo->quirka = (quirka_t *) calloc(1, sizeof(quirka_t))) == NULL) {
*error = ENDORIAN;
goto FAIL2;
}
if ((foo->quirka->fleeg = (fleeg_t *) calloc(1, sizeof(fleeg_t))) == NULL) {
*error = ENDORIAN;
goto FAIL3;
}
.
return foo;
FAIL3:
free(foo->quirka->fleeg);
return NULL;
FAIL2:
free(foo->quirka);
return NULL;
FAIL1:
free(foo);
return NULL;
}
With nested structures like these, this structured approach just seems
cleaner to me than all those temporaries kicking about.
Ben
Re: Bad C Source (Re: gzipping your websites WINRAR 40 days trial)
On Thu, Sep 04, 2003 at 03:07:08PM +0100, Lusercop wrote:
> On Thu, Sep 04, 2003 at 01:40:08PM +0100, Ben wrote:
> > return foo;
> >
> > FAIL3:
> > free(foo->quirka->fleeg);
> > return NULL;
> > FAIL2:
> > free(foo->quirka);
> > return NULL;
> > FAIL1:
> > free(foo);
> > return NULL;
> > }
> >
> > With nested structures like these, this structured approach just seems
> > cleaner to me than all those temporaries kicking about.
>
> what's wrong with:
>
> | if(foo) {
> | if(foo->quirka) {
> | free(foo->quirka->fleeg);
> | }
> | free(foo->quirka);
> | }
> | free(foo);
>
> In the error condition?
>
> or even just:
> | free_foo(foo);
> ( :-) )
>
> If you make sure your structures don't have dangling pointers, then that has
> got to be a good thing in my opinion.
>
> Still neater (IMO) than the FAIL1: FAIL2: FAIL3: approach.
Foxen for boxen. We've had variants of this conversation before. You know I
like to write my code on the assumption that the people reading it
(assuming such beasts exist) may be more inexperienced or more junior
than I am. To my mind, FAIL1: 2: 3: helps to hammer the message home.
Given that this is a fairly structured style of writing a language which
enforces so little, I don't consider this a Bad Thing.
It doesn't really matter and beyond a certain level it's a matter of
personal preference, feeling and aesthetic anyway.
Ben
Re: Bad C Source (Re: gzipping your websites WINRAR 40 days trial)
On Fri, Sep 05, 2003 at 09:46:47AM +0100, Simon Wistow wrote: > On Thu, Sep 04, 2003 at 06:44:25PM +0100, Phil Lanch said: > > On Thu, Sep 04, 2003 at 03:40:18PM +0100, David Cantrell wrote: > > > It's just this sort of thing that makes me lurve perl. > > > > you mistyped "C++". > > Without getting into a flamewar, and whilst appreciating the benefits of > compile time generic object type checking, I thought you couldn't rely > on a given C++ implementation to do auto-destruction. It's personal preference, of course, but I prefer using C to C++ for a lot of things because well, I know and understand C's edge cases well enough to either use and abuse them a bit or stay away from them. C++ seems to me to have a much larger number of edge cases - which I suppose is almost inevitable given the relative complexity of the languages. I am also probably somewhat jaded by having recently spent a large chunk of time resurrecting a large C++ application that has been slowly being written since the early 90s. It's been interesting in a lot of ways. One of them is how obvious it is that the language has changed in subtle (and not so subtle ways) in the last twelve years. The app was originally written for Sparc Solaris. I'm making it run on Intel Linux. I have neither a Sparc Solaris box, or any kind of Sun compiler. Fun. > On the other hand, IIRC, you can rely ona destructor being called. But > I think Dave's point was that he loves how Perl just does all that for > you (and therefore, I supoose, also likes that feature of Python, PHP, > Java and many more). Indeed, although the issue of *when* and *how* a destructor ends up being called depends sensitively on the garbage collection scheme, etc that a given language uses - in perl it's easy - a destructor is called immediately after it goes out of scope, because it is known to no longer have any live references to it, so this "timely destruction" is a pretty quick win. Java basically skives the issue with a "It'll get called. Don't know when, don't ask and you shouldn't worry about how it'll be called". I'm given to understand that this is quite a relevant issue to Parrot, which will be using a GC scheme which is a lot more like Java's > On a tangentially related note, I'm very rapidly starting to come to the > opinion that there are far too many applications that are written in > C/C++ which don't need to. Well, that is true, but I'm also seeing some of the problems caused by not having a (strict | anal | strong | paranoid | batshit ) type system. Certain types of bugs persist for far longer than they should in > 10 line Perl applications whereas a less laissez-faire type system would flush them out basically trivially. Whether this occupies more or less programmer time than beating your skull against a timorous typecast or other pointer pedantry will, of course, Depend. Ben
