Ok you're making some very clear and concise points here. I'll try to respond
the best I can (and try to show you're wrong at some of them ;) ).
On Monday 06 May 2002 12:27 am, you wrote:
> Hello Frank
>
> I do disagree with several of the points your making,
> however you make great points. If someone did only
> include "5" apps, well I imagine its ease would go
> right out the window. And you are right, these guys
> have spend a great deal of time, making small changes
> to packages already available, and give you some
> programs that they have created. Distros have gave
> the comunity some nice "ease of use programs" such as
> linuxconfig, diskdrake (and the many other drakes)
> among others. These guys have also done some great
> under hood changes. Red Hat has been a key respondent
> to PAM, and have improved many libraries, and xfree86,
> ive really yet to see some really improved speed
> implemenations though. If you look at programming, do
Think so? I challenge you to take xfree version 4.0, 4.1 and 4.2 and a bunch
of video cards ant test video or opengl performance. I'd swear this has
improved greatly. I think it was with 4.1 that the speed of openGL on linux
generally bypassed that on windows.
I think that the role of distributions lies, partially, in being a
contributor to free software projects - in proportion to their capacity.
Mandrake does a good job here (although I liked their old versions of
DrakConf better, at least they worked - but I digress).
The other part is maintenance of a good, consistent, coherent, up to date
collection of software; the fact a user is guaranteed that his distro will
survive (we've seen disasters happen - corel or progeny to name two) and that
he will have his security updates in a matter of a day, and his regular
updates when the distro decide they have ripened enough to be in their main
tree; and someone to answer the phone when they ring the tech support line
they paid for. It's still better to do a small project that does what it has
to do, and does it good, like ltsp. It's useless overhead to start with a
whole distro, far better to 'support' existing distro's. I fear the quality
of the current ltsp 'distribution' wil suffer from it.
Few will want just an ltsp distro. Users will get the impression that things
like ltsp can only be done with your distro then - this may be some
commercial trick, but I hate commercial tricks :) . Moreover it limits their
choice fiercely. If many users are like my case they will have a distro
installed on a powerful computer allready and then discover ltsp and install
it. (I removed the biggest hd from my p133 and used it as a thin client upon
discovering ltsp.). Do you have any data on how many of your users do have
setups with dedicated servers, how much an allready existing system and how
many of them would like to ditch their current configured system? I don't
think you can make such big decisions before having investigated this
seriously.
> you have any idea how messy it is to hard code values
> inside your program, such as a path? This is bad, and
> i mean really bad coding, linux has to put symlinks
> every which direction, just because programs arn't
> dynamic. Well, if you ever decide that is just abit
> nasty to you, look to us, we have a great deal of it
> in place. When you have an image that slows down your
> program because you wanted it beautiful, is it fair to
> the user? why have and image in kb, when it should be
> in bytes, why have all these images floating around
> when it is much more dynamic to look in a library. I
> wan't talking about implementing dll's, i was talking
> about icl's. Image Libraries, which the technology is
> already there, just nobody wanted to invest the time
> to use it.
Programming. Ok I can't program so I'll be careful with my answer ;)
First of: the hardcoding of paths and things like that in programs. I am not
sure users want as much customisations that they want to be able to change
the most basic images in programs (splash screens etc) but I think you have a
point there. It would be neat if a user could just point the program to his
own image library. I think Enlightenment 0.17 does this to some extent, by
storing themes and menu layouts in database files. About the size: I don't
understand what you mean. Is it the disk space taken up by
'rest-of-the-cluster' bytes?
> Also, distros all have their ideas, "i want something
> this way, or that", well we arn't inventing something
> to make it harder for a person. Microsoft has this
> great idea called "zero admin" well, linux is the
> furthest you can get from that. Im not going to ask
> the stupid question "is everyone so againt microsoft
> that they want linux hard to use where microsoft is
> easy?" no, it just hasn't went there yet. With use,
> to start a new service, you just put a link in a
> directory, to have an item in a menu, you just put it
> in a directory, we have in mind later a nice config
> feature, that just pulls all its info from a config
> directory, is as powerful as you make it, and unlike
> windows registry, no nasty 1 point of failure. To
> convert windows/mac/beos/* users you need it to be a
> heck of a lot easier underneath the deck before you
> going to get major changeover. You also need to try
> to get some well known apps that arn't on linux to
> join aboard.
About linux versus microsoft user-friendlyness: this is a war that has been
fought since both windows and unix co-existed on this blue ball. I'll be glad
if nobody starts trolling now, but I'll set the good example ;)
It is clear that windows and unix are two opposites from a sysadmin point of
view. Maybe the idea 'zero admin' speaks to the imagination. I think that is
mosty a lie. Generally people need the features that unix devellopers put in
really powerfull applications, like ftp and web servers, in distributions'
package management systems and so on. This, together with the security and
multi-user features of unix and the fact that open source apps are basically
develloped in wild-growth makes a pretty complicated cocktail. Granted.
Otoh, there is microsoft, who has the advantage you only have to bother about
support and updates for most of your software. This has several disadantages
(lock-in pe) but at least it looks all easy. A windows system is develloped
to make the admin or the user as much as possible away from the internals
(which gets the label'user friendlyness'). Indeed windows is not suited to
mess with internals, which seems just fine for end-users (hautainely called
'clueless end-users' by so-called linux die-hards, I plead guilty to that sin
every now and then). I don't think that this is just a point where linux (or
unix) fails because interest in a user-friendly linux system has only grown
recently. If you make something easier 'under the deck' it means making it
less powerful. The real problem is I think a double problem.
1) it is hard to configure something just superficially, often a user must
take action, read man pages, take hours of configging, especcially new users
who aren't used to this way of working. I follow a local linux news group and
it's allways gessing if this particular new user will start to love the neat
design of his shiny new linux system (it has a neat design imho) before he
gets fed up and frustrated with it and gets back to windows, becoming a
wintroll who never believes linux will ever be a serious alternative to
windows or MacOS. Afterwards they still expect the same from linux as he has
been expecting from windows so far: an easy to configure system, not a very
powerful one with lots of options, freedom, quality and general good design.
Windows allows to configure things just superficially, allways with a clean
looking gui - eventually the config may break after a while, but it sure was
easy to configure.
2) As allready stated above: wild-growth. Documentation, if any decent guides
exist in an intellegible language, take some skills to find and combine for
three different pieces of documentation to fill each other's gaps. There is
no fixed route that a user must follow to work himself through his system in
a painless way. I am 200 % sure this can in no way be fixed with a new
distro. We don't need more distro's with just an extra feature (easy
configuration of thin clients in this case). Existing distro's need to
improve. Granted that in the end your ltsp may end up as a debconf-enabled
package in a quality distro, or as an addendum in the row of *drakes. You
can't sell that I suppose...
Another point that increases the impossibility to make such a distro. It must
be able to run on different architectures. I'm pretty sure people who
currently run alpha's and sun sparcs, x86 (both ia32 and ia64!), hp machines,
and so on for application servers are an important target group. some may
even try clusters of macintoshes, They need a tested and reliable
distribution for their ltsp. And those machines cost a HELL a lot of money.
It's the reason Redhat stopped sith their sparc port at 6.2, and redhat is a
sucesful distro no? Now doesn't that set your hair on ends? :-)
As for the example about menu's I think you're underestimating the power of
linux' user-friendlyness here ;-). Your ideal sounds a lot like the system V
init scripts. Just in case you were talking about window manager menu's (I
don't use menu's because having kde's or gnome's alt-F2 shortcut,
windowmakers 'run' box or the E-run epplet for Enlightenment works faster for
me; so I didn't really study the 'internals' of this system yet): mandrake
and debian share a lot of their menu architecture. At least that's what I
conclude from meeting the name 'debian' in mandrake's man pages :-) .
Packages in debian can add menu items to /usr/lib/menu. Users can add items
to /etc/menu. The command 'updatemenu' regenerates every menu. Since the
syntax of the stuff in /etc/menu is pretty arcane, mandrake hes written a
nice graphical frontend to it: menudrake, one of the many drakes. It works
like a charm.
> I never said that i didn't have a company in the first
> place. Maybe we only have to train tech support,
> because we already have the support, oh my we also pay
> developers... scary. Also maybe another person is
Ah-ha! ;-)
What do you sell, linux support?
> right in this, that linux and unix is to broke to fix,
Unix is not broke. Not at all. We have a firm house with strong walls,
ceilings and floor, built by coopertation rather then by a central
coordination mechanism. There is just no soft sweet carpet ubder which one
can hide the crap the building is made off. Ok strong comparison but you get
the point I think :-)
> and that we should work with something else, maybe
> thats right, well see, but i think linux is healthy,
> if misdirected under the hood. Lastly above hood, it
> isnt about creating a new package manager, like
> sorcerer (they guy that said that was right, it is
> weird). Nor is it about denying people programs, we
> just dont think its right to have 3 of each in an
> initial build, if you don't like the default one, put
> another in, we'll still even suport it, 3 is just
> stupid as a default.
How do a few cd-roms of packages for each of about 5 or 6 architectures sound?
--
homepage: www.student.kuleuven.ac.be/~m9917684
jabber (=IM): [EMAIL PROTECTED]
No part of this copyright message may be reproduced, read or seen,
dead or alive or by any means, including but not limited to telepathy
without the benevolence of the author.
_______________________________________________________________
Have big pipes? SourceForge.net is looking for download mirrors. We supply
the hardware. You get the recognition. Email Us: [EMAIL PROTECTED]
_____________________________________________________________________
Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto:
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.openprojects.net
