Date: Sun, 14 Nov 2004 13:31:17 -0800 (PST) From: Doug Barton <DougB@FreeBSD.org> To: Mark Linimon <linimon@lonesome.com> Cc: ports@freebsd.org Subject: Re: portupgrade problem Message-ID: <20041114132542.N34355@qbhto.arg> In-Reply-To: <Pine.LNX.4.44.0411131332480.18539-100000@pancho> References: <Pine.LNX.4.44.0411131332480.18539-100000@pancho>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 13 Nov 2004, Mark Linimon wrote: > For some values of a and b, that is what I'm saying it would require, > yes. i.e. the only way to support the use of 'refuse' on any ports > category is to say 'this category is a leaf category and nothing in > any other category can depend on it'. > > This is what people are assuming about the 'japanese' category and > it isn't the case. We need to either change the assumption(s) or > change the behavior. Obviously the former is easier, but is it TRT? Another (and I think very useful) way to look at this would be to declare certain ports _directories_ "leaf" directories in the sense that ports whose frameworks live within them should not be depended on by other ports. Any port that is depended on by other ports should be in one of the non-leaf directories, and have the additional categories (like japanese) added as meta-categories in the Makefile. The pluses to this plan include easier refuse files for cvsup, and INDEX generation without having to download the whole tree. This could potentially lead to less load on the cvsup servers. The minuses include at least the potential of repo bloat if something grows unexpected dependencies and therefore is moved from one of the leaf directories. HTH, Doug -- This .signature sanitized for your protection
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041114132542.N34355>