a better webinterface to our packages (Was: Re: New Fedora Extras Steering Committee chair)

[seth vidal]

On Sun, 2006-01-15 at 11:14 -0500, Konstantin Ryabitsev wrote:
> À 15/01/06 10:56 AM, Thorsten Leemhuis a écrit:
> >>> I looked into repoview and while it is ok I would love to have something better
> >>> in place by the time fc5 gets around the corner. I really thing that this is
> >>> something people will like.
> >> Last time I spoke to the repoview developer, he was open to all 
> >> sorts of ideas, so you may find it easier to work with him, instead 
> >> of re-implementing his ideas all over again. :)
> > 
> > I'm sure you, andreas and everyone else interested will find a nice
> > solution. 
> 
> Well, it kinda amuses me, since web interface to packages is a 
> problem with a well-known set of constraints – both real and 
> perceived. The greatest limitation is the attempt to make it not 
> require anything on the server-side, in order for the mirrors to not 
> have to provide special scripting-enabled directories and still have 
> all the same data as in the main repository. While this is a 
> worthwhile goal, this prevents me from implementing any sort of 
> searching and other features. If pages were rendered dynamically, 
> then we could just use yum's own sqlite databases to query and 
> render pages on the fly, with lots of niceties, but I am reluctant 
> to do it, since I like the simplicity of pre-rendered pages (a lot 
> fewer bugs, a lot more secure).
> 
> I understand that repoview also adds a significant overhead to the 
> generation of large repositories (though I must note that on my 
> P-Mobile 1.4GHz the entire process of parsing and then generating 
> all pages for 3000 packages takes 80 seconds, which is pretty good 
> in my book).
> 
> What I think I can do immediately is implement a feature where the 
> program finds ALL subdirectories with repodata in them and then 
> generates one set of pages for all the architectures it finds. This 
> way instead of having to run it 4 times for SRPMS, i386, ppc, and 
> x86_64, you can run them in the /extras/4 directory and have one 
> tree listing all architectures available. Would that be useful?

Here's what I think..

I think repoview on the mirrors is great it lets anyone have a copy of
it, etc. However, we could have a static link in the mirrored data that
points to a searchable interface that lives on fedoraproject.org.

Thoughts?

Then again, I wonder how much of this is obviated by the presence of
good searching interfaces in the tools we have in the distro, now?

-sv