Jesse Keating wrote:
There has been lots of talk in the past about making it easier to
experiment with changes in Fedora, particularly when many packages are
involved. We never have a good story for this other than "use mock".
However now that we have things like
fedorapeople.org and lots of smart
people wanting to improve things, I think we can have a much better
story, one that can rival Ubuntu's offerings
(
https://help.launchpad.net/PPAQuickStart).
So from the "I don't have time to work on it, but here is what I think
we could do" department I bring you a concept of Koji Personal Repos,
or kopers (pronounced like copper, the metal).
Please have a read at
http://fedoraproject.org/wiki/JesseKeating/KojiPersonalRepos and let the comments fly.
In the arena of available buildsystems (IE: SuSE, Ubuntu, and Fedora)
from my research we are behind, very far behind. Part of this is
builders and part of this is storage. I've been looking for 20T of
storage or so (which is to include the current arch's and the secondary
archs) but if we're going to advertise personal builds and start
considering implementing other open source projects (like OLPC, Creative
Commons, etc) which we're already sort of supporting, then we need to
think long and hard about this.
We really need to put some investment into our current system, both in
terms of Koji (Read: making sure the kernal and OOo builds have a higher
weight then ytalk, continue our GC stuff, finalize secondary arch
integration) and hardware. Getting enough people to make enough noise
so our blade center gets installed (it's coming up on a year of paid
service for a paper weight sitting on the floor). There's also storage,
the estimates on secondary arch's are in (note: thanks to Debian we're
already way behind Ubuntu on this -
http://cdimage.ubuntu.com/ports/releases/gutsy/release/ ) But estimates
on personal builds are still a question mark. I'd like to find
somewhere around 30T for this and dedicate it JUST to the build system.
I'd hope this would last us for the next 3 -5 years.
I also think we'll need to find another primary mirror somewhere for
distribution. We've had some queries for distribution already (Creative
Commons). If hosted takes off as I think it might, we'll need another
distribution medium and it needs to not be our build system. We could
upgrade our primary mirror but that's complicated because it is 3
netapps spread across multiple locations, we'd have to upgrade them all
and we'd be under the mercy of Red Hat with regards to major releases so
we wouldn't be in control of our own destiny and that's really really
bad. We have torrent, but torrent is ill suited to distribute some
things, like nightly builds or updates-released repositories for our
secondary archs. Also our new torrent site will have a bandwidth cap
initially.
I've been trying to get an actual budget for our Infrastructure, I
submitted a budget to be approved a few months ago but have not heard
back (and likely won't for a month or so). Even if it does get
approved, it does not include the above requirements. I'll create
something separate document for what it would take to bring Fedora to
the next step in terms of Buildsystem and the next step in terms of
being able to better support non Fedora OS endeavors (distribution, etc).
Any thoughts, questions, comments?
-Mike