It seems that what we're talking about here is a matter of standardization, not of storing Group info in comps.xml being better than storing it in the package header. I mean, someone could create a comps.xml with wacky groupings just as easily as they could add wacky groupings to their package's headers, right? Or am I misunderstanding some aspect of this? I concede the point about utils like anaconda being geared more toward using comps.xml than the Group field and agree that we should settle on one rather than both. But I'm not convinced that it's better to keep all this information in one file (even one file per repo) instead of in the packages themselves. What, other than current development trends, warrants the use of a file that would need to be updated every time a package got added to a repository if reaching an accepted standard for Group field values would suffice? --Brad
