Michael Orlitzky <mich...@orlitzky.com> writes: > On 03/01/12 23:43, Keshav Kini wrote: >> I don't understand. Why would it be *faster* to do version bumps if >> sage-on-gentoo gets into Gentoo proper? Overlays are always more nimble >> than the Gentoo tree, as far as I can see. > > If we're to distribute sage via source, we need some way for users to > build all of the prerequisites. I think prefix is the way to do that, > and prefix uses the portage tree, so it's important that nothing in the > portage tree breaks sage. > > If sage-4.8 is in the official portage tree, then other Gentoo devs > won't commit something that breaks it. For example, if sage-4.8 depends > on sys-devel/gcc=4.5.3, the gcc maintainer won't just remove that > version while sage-4.8 is in portage.
I see what you're saying. But if we packaged Sage with prefix, I'm sure we'd prioritize our own ebuilds over those from Gentoo's portage tree, and make sure to have every dependency of Sage in our own tree of ebuilds. With strict dependencies in our ebuilds, new versions in Gentoo's portage tree would not affect the building of Sage, as it would still select the ebuilds from Sage's tree when it was trying to build dependencies; neither would old versions being deleted from Gentoo's portage tree, as the old ebuilds would still exist in our own Sage portage tree. Take a look at lmonade, for example - it is Prefix-based, yet it does not contain a copy of Gentoo's portage and does not sync to it, afaik (though I imagine this will be added later as an option). >> sage-on-gentoo being an overlay also makes it easier to create a Prefix >> distribution for Sage, I think. Eventually the Prefix distribution's >> portage tree would be the main effort, and sage-on-gentoo would be an >> adaptation of it as an overlay for users who are running Gentoo on the >> top level of their system. >> >> Making sage-on-gentoo a part of Gentoo proper ties it too strongly to >> the Gentoo distribution, IMO. Certainly it is a good thing to push as >> many ebuilds from sage-on-gentoo to Gentoo as possible, but we should >> still maintain our own collection of ebuilds for Sage if we want to go >> forward with Prefix, no? > > I think we would do all of our development on github in an overlay, but > then commit releases back to portage. That would ensure that all of our > dependencies keep working for the lifetime of the release. Again, we should ship our overlay with the distribution. We should of course try to push from our overlay into Gentoo as well, but for non-Gentoo users it is a waste of time to delay a Sage release because we don't have the correct ebuilds in Gentoo's portage, for example. The Gentoo portage tree would only serve as a repository from which to install other applications into your Sage directory, such as what we now call optional and experimental SPKGs, and a plethora of other packages that Gentoo users enjoy the use of. And even some optional and experimental packages we might need to supply ourselves, if they are too niche to be in Gentoo's portage. My point is that Gentoo should not be the source of anything we now call standard SPKGs - this ties Sage to Gentoo too strongly. -Keshav ---- Join us in #sagemath on irc.freenode.net ! -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
