Re: GNUstep repository (was LinuxSTEP + Integration of apps)

[Tim Harrison]

(pardon the cross post to the LinuxSTEP-General list -- I figured people 
might be interested there, as well)


Dennis Leeuw wrote:

> The problem of LinuxSTEP vs SimplyGNUstep is the different approach.

I don't see this as a problem.  I see it as just a different approach.

> - Create a small and simple distribution meaning with only the bare
> essentials:
>         - libc (in /lib)
>         - basic Un*x tools (for /bin and /sbin) think of login, awk, sh, etc..
>         - only the absolutly needed libraries (in /usr/lib)
>                 like libxml, libtiff, libjpeg, etc.
>         - the un*x build tools (in /usr/bin and /usr/sbin)
>                 like make, gcc, etc.
>
> - X11 living in /usr/X11
>
> - GNUstep living in /usr/GNUstep

Then you've negated one of the major reasons for LinuxSTEP to exist, and 
hence, there's no point in proposing a merger.  If you want the 
traditional FHS structure, with GNUstep in /usr, then use SimplyGNUstep. 
 Or Debian.  Or Slackware.  For SuSE.  Or Red Hat.  However, that's 
contrary to what we're doing.

> - The initial problem probably lies in the packaging and a good boot
> process (although I am in the middle of rewriting all rc.d and init.d
> scripts and creating a single tree of basic system configuration, thus I
> know that it is doable, a package manager is an entirely different
> subject. I have started over and over again on this subject. Package
> management on Un*x is a hell... IMHO.)

That wasn't our initial problem.  Our initial problem was the FHS.  We 
didn't necessarily believe it was the best way to go.  So we took a 
different route.  Some people might disagree, or feel that we're not 
doing it "right".  That's fine.  Not everyone likes everything.

Packaging was also a problem, but Eric's LSPM system is bloody fantastic 
(IMHO), and solves, at this point, about 90% of the problems I've ever 
had with package management.

As for the boot process, it's something I've been working on as well. 
Yes, it's a pain, but basically, you can do anything you want with 
bootup, once the kernel has passed control to init (which can be a 
binary of your own design).

> I think with this minimalistic setup one should be able to create a
> single CD with a life system incl. the sources (almost as simple as
> SimplyGNUstep).
>
> An additional CD could hold the Developer apps (incl. Sources) and a
> third CD could be used to hold Postgresql with GDL and WebObjects.

Why would one need all those CDs?  You can do a very basic live 
filesystem, include all the developer tools, and database stuff on one 
ISO.  Don't forget, the majority of the world is still on dialup, so 
downloading three ISOs is a pain for some.  Plus, inherently, three ISOs 
 is annoying.  Plunk it all onto one, and it's just simpler.  No 
swapping CDs when installing, also.

> I think a team of about five to ten people could pull this off. To

If following the FHS, and no porting is required, you wouldn't need five 
to ten people.  The main reason we're taking so long to get our stuff 
out the door is porting.  We need to make things work with the structure 
we've defined.

> create this solution one needs a server that holds all the data, and
> which could also serve as the repository for the GNUstep apps. A single
> maintainer would be able to package from the different sources the
> sources of the apps and provide them to the GNUstep community.

I don't think one needs to centralise everything for a solution. 
GNUstep.net hasn't centralised all the applications, nor have any of the 
other GNUstep-specific application sites.  Most people release their 
projects on their own web sites, and just send out an email.  They tend 
not to want to go through the process of signing into someone else's 
server, through another web interface, and uploading their latest 
versions there, or to any number of other sites.

With the type of FHS system you're proposing, it sounds to me like 
things wouldn't need to be specific to the system.  That's one of the 
benefits of GNUstep, as I see it.  So, the only difference is packaging. 
 If there was a packaging system accepted for general use, it would 
most likely be something provided via the makefile system, so one could 
type "make package" or somesuch, and *poof*, you have a package.  No 
maintainer needed.  Just an extra command by anyone who's releasing.

Of course, this means that GNUstep maintainers would need to accept some 
submissions into -make from those making packaging systems, so that one 
could potentially type "make lsp", for example, and create an LSPM 
package.  Same for any other package management system.

> What is needed is a single server to start with. Enough bandwidth for
> the ISO downloads and some spare time and dedication.
> So the first question that pops up is where do we get the money to
> install a server with enough bandwidth.

Where I work, bandwidth is not a limitation.  Nor is having a box (or a 
ew boxen) in the data centre a problem.

> And the second question is, are there enough people willing to invest
> time to pull this off.

I think you missed the first question.  Are the LinuxSTEP and 
SimplyGNUstep camps willing to/able to merge.

In my opinion, the goals of the individual projects are potentially too 
far apart.  There would have to be a fundamental shift from both camps 
for a merger to occur.



--

Tim Harrison
tim@linuxstep.org
http://www.linuxstep.org/