Shawn Walker wrote: > 2008/6/3 <[EMAIL PROTECTED]>: > >>> However, one problem with this is that it creates a situation where >>> you are publishing a package that depends on a file in a package that >>> hasn't been published yet. >>> >> How is this any different from the behavior that we have today? I can >> write a package with a dependency on a package that isn't in the >> repository. At least in the current code, we don't require that that the >> dependent packages be present in the catalog. >> > > It is different because the server won't have the information > necessary to tell you which package provides the file. > Nor client do. Think about third-party repositories. mutt depends on /usr/lib/sendmail, only sendmail and postfix are in official repository. However, mutt would be satisfied with exim (or even nullmailer), which also provides /usr/lib/sendmail -- and maybe someone else packaged exim (personally I prefer it over postfix and sendmail)
With virtual packages that's not a problem at all: mutt requires local-mta; every package claiming he can replace say sendmail may include something like "provides local-mta" in metadata. >> We're talking about a mechanism that allows the developer to >> specify a file name, and if possible, get that translated to a package >> dependency. This needs to happen at publication time; however, it would >> make more sense to have the publication client (pkgsend, solaris.py, >> whatever), and not the server, figure out the dependency. >> > > Yes, that is what I was pondering. > > _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
