Date: Tue, 24 Aug 2004 14:34:34 -0500 From: "Brandon D. Valentine" <brandon@dvalentine.com> To: Oliver Eikemeier <eikemeier@fillmore-labs.com> Cc: Kris Kennaway <kris@obsecurity.org> Subject: Re: LATEST_LINK unique or not? Message-ID: <20040824193434.GC86834@brandon.dvalentine.com> In-Reply-To: <B932C907-F5C1-11D8-8CAA-00039312D914@fillmore-labs.com> References: <412AEBA6.17012.839357E2@localhost> <B932C907-F5C1-11D8-8CAA-00039312D914@fillmore-labs.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Aug 24, 2004 at 01:35:39PM +0200, Oliver Eikemeier wrote: > FWIIW, the real `fix' would be to require uniqueness of LATEST_LINK, > even when NO_LATEST_LINK is set. I think we have more than one use for a > unique package name without version number. Should I just make a patch > for the tree? > > As said above: I think a global unique LATEST_LINK is beneficial, and > since we already have something like this in CVSROOT-ports/modules, it > shouldn't be too difficult. I think this is a great idea and appreciated Kris's crackdown on it several months ago, even setting LATEST_LINK for my ports that don't actually build packages, thus never using LATEST_LINK, with the anticipation that LATEST_LINK is becoming a defacto unique identifier. So let me cheer you on here if you're suggesting that portmgr officially require all ports to have a globally unique name. I've always been a little uncomfortable with the name LATEST_LINK for the globally unique identifier of a package though. There is a UNIQUENAME in ports right now, used currently to define OPTIONSFILE, but nowhere else. LATEST_LINK is currently only used to define UNIQUENAME and then in targets which deal directly with the creation of actual package Latest links. I understand why these two variables interrelate in this way given how long LATEST_LINK has been around and how briefly UNIQUENAME has been around. If portmgr is going to formally encourage unique package identifiers though, I'd like to ask that they also consider reversing this relationship. In other words, for ports which don't set UNIQUENAME, but do set LATEST_LINK, i.e. every port currently in the tree, set UNIQUENAME to LATEST_LINK. Going forward discourage porters from using LATEST_LINK in favor of UNIQUENAME, and for ports which define UNIQUENAME but not LATEST_LINK, set LATEST_LINK to UNIQUENAME. This way, if in the future if you want to change the usage of LATEST_LINK or the architecture of the package repository the mechanisms used to create the PKGLATESTREPOSITORY are independent of the unique identifier of the port, which should almost never change. It might even be desirable at some point to do an s/LATEST_LINK/UNIQUENAME/ on the ports Makefiles. Aesthetic concern I know, and LATEST_LINK certainly works, but you might make the purpose behind setting this variable more obvious to porters if it was called UNIQUENAME and you would detangle the PKGLATESTREPOSITORY magic from the UNIQUENAME, which really shouldn't be tied up in it. Brandon D. Valentine -- brandon@dvalentine.com http://www.geekpunk.net Pseudo-Random Googlism: beer is a labor of love for little snoqualmie brewery
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040824193434.GC86834>