Re: Repo information inside rpm?
Morten Kjeldgaard <[email protected]> Fri, 13 Aug 2004 15:33:20 +0200 (CEST)
| Newsgroups | gmane.linux.redhat.rpm.atrpms.repo-coordination |
|---|---|
| Message-ID | <[email protected]> |
>> IMO, if you're going to pack this in the RPM using an existing tag, a better >> target would be the >> Packager: >> tag. > > Well, I think Morten's reasons for doing this is that the Distribution tag > is not displayed in the rpm -qi output. And that's what I probably want > too if the decision is to not have it human readable :) Exactly! And the Packager tag is in fact pretty useful in that it tells you the name of who in the world created the package (and possibly signed it). > I would also prefer a scheme that doesn't require a specific position, > something like: > > Vendor: name=rpmforge, url=http://rpmforge.net/, repo=dag, dist=fc2 > > Or use the different headers for each type of information: > > Distribution: repo=dag, dist=fc2 > Vendor: name=rpmforge, url=http://rpmforge.net/ > Packager: name=Dag Wieers, [email protected], url=http://dag.wieers.com/ I briefly considered the key-worded approach, and it has several advantages. Primarily, it is much more flexible, and it is easy to add new keywords as time goes on. The disadvantage of using more than one tag is that it will be very difficult to maintain a standard, and the parser of the information needs to do a lot of error-checking: which keyword belongs to which tag? And can you generate a sources.list entry from the information? I think that is pretty crucial. So I definetly think we should stick to using just one tag. I like Distribution: but it could be Vendor:, which is not really used by anybody. I don't care that much, it is more important that we come to an agreement on something. My proposal has the advantage that it is simple, short and sweet, and can be cast into rpm macros. > Vendor: name=rpmforge, url=http://rpmforge.net/, repo=dag, dist=fc2 Perhaps a minor thing, but "fc2" is what I call "release". "fedora" would be the distribution, "fc2" the release. (And you can't regenerate the sources.list line from your example). > 1. I wouldn't use a URL to identify a repository. apt.sw.be is temporary > for me and I don't want that in any of my packages anyway. Hmmm. But that is your primary distribution site (right?) and presumably, if you move your stuff somewhere else, the package that has this info in it would still be available from there. I think the URL has to the same as goes in sources.list or yum.conf (so you can generate them from the RPM). Cheers, /Morten