On Sat, 30 Jul 2005, Kito wrote:

> 
> On Jul 30, 2005, at 8:12 PM, Hasan Khalil wrote:
> 
> >As of yet, portage is not suited for what we have now termed 'pathspec', or
> >'installing to an alternate prefix', or 'using portage as a secondary package
> >manager', etc. Changes are being made to portage that will allow for features
> >like this, and should be included in the next major release (some months away
> >still).
> >
> 
[snip]

> The main problem as I see it, is you have a live tree of some 
> ~10,000 linux based packages, with a userbase of >100,000 LINUX users, a 
> dev team of >350 linux developers...how on earth do you convince these 
> linux users and devs that a massively huge project like supporting 
> arbitrary install prefixes is worth the trouble, especially when it 
> would mostly benefit a sideproject with 3 devs and probably only slighty 
> more users?

Remember that for some of us, it doesn't matter if no more than a tiny 
fraction of ebuilds work. One should not confuse Gentoo (i.e. the portage 
tree) with Portage itself. Non-Gentoo developers, distros and O/S's can 
benefit from a portable portage, even if it comes with an empty portage 
tree. By "portable portage", I mean that it would support new ebuilds that 
will play nicely on arbitrary host.

But the question remains, how to bring the existing ebuilds along for the 
ride? Kito is right that most linux devs aren't going to care too much. 
Most of them are not in a position to test their ebuilds on half a dozen 
different platforms. But then, they don't all test on half a dozen 
different linux architectures anyway.

Hasan, you mentioned pathspec and prefixed installs, and future portage 
features to accomodate these. Is there more information available anywhere 
on the portage roadmap and the particular future portage features you are 
referring to?

-f
-- 
gentoo-osx@g.o mailing list