Re: RFC: implementation of package pools
Wichert Akkerman wrote:
>
> No it would not, that's simply not the purpose of unstable. Unstable
> is where the latest and greatest and most broken lives.
Okay, so can I upload a program that doesn't work at all to unstable?
(When I become a real developer). What happens when that's a very important
component? I think the "most broken" must have some limit.
> > With the kind of testing I suggest, it isn't.
>
> You're still not understanding the fact that package pools has *nothing*
> to do with your testing. Package pools are solely about how to install
> a package in an archive, how to make it available to users, etc. What
> you install and the tests that might accompany that are a different
> issue.
I know what package pools are. I'm saying that it should be developed
with support for testsuites per package in mind! dejagnu anyone? Of
course the development of testsuites will be orthogonal to package
pools. Testing should be a separate module, I already said that. Don't
ignore that there is a relation between the archive and testing though.
The interface between them should be fairly easy though, but if specified
earlier, more packages can take advantage of it.
My point is this:
specify an interface for test suites that the package developers
themselves write. for their own convenience. in any case, they can write
that themselves, but formalizing it will encourage them!
> distributed OSes as a different beast, and the problems are different.
> You might want to look at http://www.globalfilesystem.org/ .
>
Thanks for the information ;) Didn't I already tell you that I read
a lot of distributed fs papers and that my area of research was parallel
programming? You really have no trust in CS grad. students ;)
Of course GFS was among them. It's suited for beowulf setups, but I didn't
find it general enough. I liked Zebra a lot, but performance isn't
the only issue "serverless" operation is a must, and scalability and
security...
[not mentioning other mandatory things like location transparency]
Writing a distributed fs isn't that different for a single machine,
since when you run it on a single machine it should give decent
performance. LFS would be nice on Linux, but I hope progeny is already
implementing it!
> > If you use it the correct way then it does.
>
> I don't agree with that; in the end this is personal taste.
>
I agree that it's personal taste. I'm an asm programmer originally,
so I never found that much satisfaction in C :) C++ gives me what
I expect from a higher level language. I'd use a functional PL if
I weren't worried about performance.
> > You haven't answered to my request on the dpkg.c Could you please
> > give a brief overview of the code and algorithm used for satisfaction
> > in checkinstallable? I'd really like to review it and contribute
> > to the code. Is there a well commented version of it somewhere?
>
> I didn't write or even look at dpkg.c, so I'm the wrong person to
> ask.
Anthony must know about it, perhaps I can consult to him in private.
Thanks,
--
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: erayo@cs.bilkent.edu.tr
www: http://www.cs.bilkent.edu.tr/~erayo
Reply to: