Re: [9fans] s3venti
and an issue related to the fact that we need to encrypt users' data. For the record, s3venti does encrypt blocks that it writes to S3. It uses a single key, making it rather vulnerable to dictionary attacks, but I haven't come up with a way to do better without changing the venti protocol. Suggestions are welcome.
[9fans] s3venti
I mentioned in passing some time ago that I was working on a venti server that uses Amazon S3 as a storage backend. There is now code in /n/sources/contrib/rcbilson/s3venti . Beware sharp edges. I have pumped a fair amount of test data through it successfully, but I wouldn't recommend trusting anything important to it yet. There is a man page. I started writing it under plan9, but for irrelevant reasons later switched to plan9port, so that's where it's known to work (on Linux, at least). I would hope and expect that moving it back to native plan9 would be a small job. Questions and comments are welcome.
Re: [9fans] plan9.bell-labs.com
Venti arenas stored on S3, with locally hosted indexes? You can't just throw a whole arena into an S3 file, since the S3 API doesn't support seeking within files. On the other hand, an S3 bucket is essentially a big hash table, and so is venti. So I opted for the simple approach of storing the blocks themselves in S3. Given that, a local index doesn't help much. A local bloom filter probably would help, but for now my server is stateless (aside from cache). Right now the server supports the basic venti api, but it falls over if anything goes wrong on the connection. Once I make it more robust, I'll put it in contrib for interested parties.
Re: [9fans] plan9.bell-labs.com
On 9/20/07, David Leimbach [EMAIL PROTECTED] wrote: And I just had a really evil idea involving Amazon's S3 storage service... I think we may have had the same idea. And I have code, too.
Re: [9fans] Plan 9 household shared file server
Fastest way I could get this out in a pinch was to use a screenshot :-) Wild guess: aquarela must be run as bootes so that it can assume the identity of the user logging in. As I recall, it's not really prepared to deal with any other circumstance. What Steve said regarding share names, plus: if you browse your machine, you'll find a single share named local by default. This will contain whatever is under /n/local. In the default installation there's nothing there, but you can put things under there or (better, I think) use /lib/namespace to bind things there.
Re: [9fans] OT: cannonical set of queue ops
Are we suffering from a lack of vision, or is some way of checking the number of items on a queue inevitable? The problem with these sorts of functions is that the results that they give can't, in general, be used for much. For instance, if IsQueueEmpty returns true, that answer isn't worth very much if, by the time you act on that information, some other thread could have added an item to the queue. They're dangerous functions to have around, since often people will be tempted to use them inappropriately, relying on the answer even where the potential exists for other threads to perform concurrent operations on the same queue. But there are certainly cases where their use is appropriate, and it sounds like you have one of those cases here (assuming no thread or interrupt handler is going to add more items to the queue as it drains). In summary: it can be useful to have such a function, but it probably won't (and shouldn't) be used often.
Re: [9fans] OT: cannonical set of queue ops
On 12/6/06, LiteStar numnums [EMAIL PROTECTED] wrote: For instance, if IsQueueEmpty returns true, that answer isn't worth very much if, by the time you act on that information, some other thread could have added an item to the queue. Wouldn't it be natural for such a function to block the active thread until the desired result is achieved? Since this clean up thread wouldn't really need to be on a hard deadline, wouldn't WaitQueueEmpty make more sense then a IsQueueEmpty? The writer may be more difficult, but again, if the Queue is full, it must wait anyway, so blocking here wouldn't be too problematic, methinks. You could do that as well, but that doesn't address the fundamental problem that I was pointing out regarding IsQueueEmpty: once you wake up you don't really know anything about the state of the queue, unless you are sure that no other thread could be adding items. All you know in general is that the queue was recently empty. I agree with you that it's usually better to block than to busy-wait.
Re: [9fans] auth stuff: never saw this one before:
On 12/4/06, ron minnich [EMAIL PROTECTED] wrote: Uh, hints ? On 7/9/06, Russ Cox [EMAIL PROTECTED] wrote: It means that the bootes (whatever your hostowner is named) key in keyfs (/mnt/keys) does not match the one you typed into the nvram that the server factotum is using. For what it's worth, I did two installs this summer and wound up with this error both times. I didn't think I was *that* bad at keeping track of passwords... Both times it worked after I nuked the nvram.
Re: [9fans] 9pm/Plan 9 kernel in user space
On 10/19/06, Sascha Retzki [EMAIL PROTECTED] wrote: I can comment on that. Getting the facts, Microsoft once wrote some API so that anyone, with the right SDK, could write virtual file system drivers - they actually presented that stuff via a ftpfs, which made me optimistic that, if one gets this SDK, 9p4win32 could be written. It seems like if you already have access to the Windows DDK, then http://www.acc.umu.se/~bosse/ has the rest of the pieces needed to do such a thing, and examples. Of course, I can't confirm that it actually works. I took a look at the examples and decided that I didn't have the stomach for it.
Re: [9fans] safe to pull from sources now?
On 9/25/06, David Leimbach [EMAIL PROTECTED] wrote: Just wondering if anyone has had success with this since the last hardware difficulties that were experienced. I pulled about a week ago and things seemed sane. No obvious problems since.
Re: [9fans] p9p for win32 ?
Anyone done any work trying to port p9p to win32? Actually, I've been trying to get around to that, but I haven't gone any farther than browsing around Russ's for reference only win32 code. I am after just the command line tools (ls, cp, awk, rc, mk etc) though the full gui would be nice :-) I want vac, although the rest is gravy. If not I may make a start, using mingw as the compiler, unless anyone has strong views whi I should use MSVC, Borland, LCC etc. Russ's code seems to be written for MSVC, although I don't see it being a big issue either way. I was going to use mingw myself, for what it's worth. - Richard
Re: Re: Re: [9fans] Aquarela usage
On 9/5/06, David Leimbach [EMAIL PROTECTED] wrote: On 9/5/06, Richard Miller [EMAIL PROTECTED] wrote: How does aquarela authenticate then? Is it the user that started the aquarela server or any valid plan 9 user? grep -n auth_ /sys/src/cmd/aquarela/*.c Looks like it should work for any user with a key for proto=mschap. Yep, I don't think I have any users with said key, nor am I immediately aware of how to add that. But I'm sure I can find out. I've never had to add any special keys in order to access an aquarela share from a Windows machine -- I can connect to the share using the same password I use for drawterm. I have also never had a need for aquarela's -n flag.
Re: Re: Re: Re: [9fans] Aquarela usage
On 9/5/06, David Leimbach [EMAIL PROTECTED] wrote: On 9/5/06, Richard Bilson [EMAIL PROTECTED] wrote: I've never had to add any special keys in order to access an aquarela share from a Windows machine -- I can connect to the share using the same password I use for drawterm. I've been trying that, drawterm works, aquarela isn't working. I have also never had a need for aquarela's -n flag. I'm trying to connect to the local filesystem using \\ip\local or smb://ip/local. So far no luck. Do I have to do something special to /n in the namespace I launch aquarela in? Dave If your problem is authentication, no. But if your credentials are accepted, my message from July 18th might be relevant: On 7/14/06, Richard Bilson [EMAIL PROTECTED] wrote: I assume that \\host\local is supposed to refer to some namespace constructed for the authenticated user. Probably it should contain files and such. How do I achieve this? Ok, I figured this out. \\host\local refers to /n/local. But, by default, there is nothing mounted at or bound to /n/local. Putting bind / /n/local into /lib/namespace.local gives me the behavior I expected. I was going to find somewhere in the wiki to put this, but I'm still not completely sure that this is the way it's meant to work. Of course, it's nice and flexible in the sense that you can create a custom namespace for use by smb clients, but it ought to be documented. One way to make sure that the authentication is working is to browse \\host\sources or \\host\dump, or another share name corresponding to a 9fs argument.
Re: [9fans] vmware?
On 8/4/06, LluĂs Batlle i Rossell [EMAIL PROTECTED] wrote: Anssi Porttikivi wrote: I can not find a Plan 9 disk image for vmware. Is there? Or do I have to make my own? But I think I can not do that with the free vmware palyer? I think Vmware Server got free - maybe it helps, although I have not tried it. I have installed Plan 9 on VMware server and it works fine except for the graphics. That didn't bother me, but might affect other people's decisions. The graphics seem almost correct, but certain things don't get drawn, which results in it being unusable.
[9fans] Re: aquarela
On 7/14/06, Richard Bilson [EMAIL PROTECTED] wrote: I assume that \\host\local is supposed to refer to some namespace constructed for the authenticated user. Probably it should contain files and such. How do I achieve this? Ok, I figured this out. \\host\local refers to /n/local. But, by default, there is nothing mounted at or bound to /n/local. Putting bind / /n/local into /lib/namespace.local gives me the behavior I expected.
[9fans] aquarela
I'm having a problem with aquarela which may be related to my tenuous grasp of the authentication mechanisms. I can open \\host, authenticating using my username and password on host, and see local. I can open \\host\local, but I see no files and I can't create any. I can open \\host\sources and from there I can browse sources. I assume that \\host\local is supposed to refer to some namespace constructed for the authenticated user. Probably it should contain files and such. How do I achieve this?
Re: Re: [9fans] Which thing was harder for u to grasp relating Plan 9?
On 7/12/06, Francisco J Ballesteros [EMAIL PROTECTED] wrote: Going back to my question about the top-10, I remember that it was very hard to me to understand why bind / / was there after I did a bind -b whatever / Now it seems clear and very simple, but it was not at all for me initially. For me, the whole idea of bind confused me, because I expected more from it -- I thought it would be more like a BSD-style union mount. It took a while for me to realize that it's actually a much simpler mechanism designed to solve a different problem.
Re: [9fans] New cpu server woes
On 7/9/06, Russ Cox [EMAIL PROTECTED] wrote: It means that the bootes (whatever your hostowner is named) key in keyfs (/mnt/keys) does not match the one you typed into the nvram that the server factotum is using. Perfect. It works now. Thanks, Russ.
Re: [9fans] OT: plan9port compilation failure
On 7/6/06, ISHWAR RATTAN [EMAIL PROTECTED] wrote: Did a fresh cvs checkout and tried to compile. System is (via uname -a): SunOS cps222 5.10 Generic_118833-03 sun4u sparc SUNW,Sun-Blade-100 In plan9 directory (via ./INSTALL): ... cd /home/user/rattan/plan9/src/libthread; mk all 9c -I. pthread.c sparc-ucontext.h:20: error: conflicting types for 'makecontext' /usr/include/ucontext.h:43: error: previous declaration of 'makecontext' was here There's a block of code that is #ifdef __sun__ in threadimpl.h that breaks Solaris. I was just in the process of preparing a diff when I was distracted by other things. In particular, this block is unnecessary and harmful: #if defined(__sun__) # define mcontext libthread_mcontext # define mcontext_t libthread_mcontext_t # define ucontext libthread_ucontext # define ucontext_t libthread_ucontext_t # include sparc-ucontext.h #endif Having eliminated that, however, it's working well for me.
[9fans] New cpu server woes
So what does one do when drawterm tells one that ?you and auth server agree on password but ?server is confused cpu: server lies got 533e77d4931126f4.67665283 want 79f86876191c41ff.0: cs gave empty translation list In this case, the auth server and the cpu server are one and the same, newly configured following the wiki instructions (first time for me, so don't discount the possibility of something truly brainless). The system was previously configured according to the drawterm to your terminal instructions, although besides that not a lot has happened to this installation. The cs error is probably trying to tell me something, but ndb/csquery gives the right answer to every question I ask when run from the console. Thanks, Richard
[9fans] Re: LinuxTAG
On 4/24/06, erik quanstrom [EMAIL PROTECTED] wrote: der geist is willig; es fehlt mir am geld! - erik On Mon Apr 24 21:05:27 CDT 2006, [EMAIL PROTECTED] wrote: Guys, is anyone going to be at LinuxTAG this year ? http://www.linuxtag.org Thanks, Roman. P.S. Since I'm on the speakers list I can help with getting passes to the showfloor and such...
[9fans] Re: Good enough approximation for ape/pcc
On 4/11/06, Charles Forsyth [EMAIL PROTECTED] wrote: the problem is that the X11 libraries are required for p9p to be useful. the X11 libraries are probably not so bad (well, not in that sense), but i wasn't referring to using ?[acl] for p9p anyway
Re: [9fans] 9pm customization
Getting to the point, for those of you who still use 9pm, what have you added to your path to make it more livable? Have you found that any of the utilities that ship with Inferno are more up to date? I replace srx.exe with PuTTY's plink.exe, giving me support for SSHv2 and PuTTY's authentication agent. As a result, I can use sam -r to edit files on UNIX machines relatively comfortably.
Re: [9fans] A Plan 9 C request....
On 3/1/06, Skip Tavakkolian [EMAIL PROTECTED] wrote: when it was added to C++, i felt that the scope of 'i' wasn't natural; it goes beyond 'for's closure. i like a behavior like this: { int i; for (i = 0, ...) ...; } This (the way you like it) is the way it was eventually standardized in C++. Any modern compiler that I have used does it this way.
Re: [9fans] p9p: SAM snarf with X
On 8/28/05, Russ Cox [EMAIL PROTECTED] wrote: to add chording (presumably cording has something to do with processing /sys/log), there is some code that is marked with if(chording) in the plan9port version of samterm, but rob got it to hang so i disabled it. I've used the chording support in sam without incident (which, of course, proves nothing about race conditions), but I should warn that it does two things that annoy me, given that I'm used to the Acme way: - B!-2-3 (cut + paste) makes a clean file dirty - you can't double click to select a word and then chord the resulting selection I haven't looked closely at the problems, but my intuition says that they're both hard, given the sam/samterm split. - Richard
Re: [9fans] Absent friends of Boyd list so far ...
On 7/19/05, sam ducksworth [EMAIL PROTECTED] wrote: please add me to the list. i didn't know him personally but i always enjoyed reading his posts on 9fans. Likewise for me. I'm sure he wouldn't know me from Adam, but I always enjoyed reading his take on things. - Richard
Re: [9fans] acme/sam junky in need of advice
On 5/9/05, Richard Bilson [EMAIL PROTECTED] wrote: As I recall, attempting to use any sort of shell function fails. But I haven't tested this since SP2, so you never know. I'll try it again tonight, and send you an example if it still fails. It seems to work now, mysteriously. Oh happy day!