Metadata: Since we need/want programmable I doubt we'll get disagreement on XML. Static or dynamic is more interesting.
Questions: 1) Do we need server side metadata? I suspect yes, for referencing the 'version' of the repository spec being followed, if nothing else. That said, I could imaging "owner" (human), mailing alias, and lots of admin related information that tools might work with. 2) Do we need per group metadata? I suspect yes, for quirks. I could think of a 'manifest' describing what is there, e.g. for Java -- stating the package for this project, etc. I could see this being human maintained, or tooling. 3) Do we need per version metadata? Hmmm... :( I guess it's doable, tools could easily provide. The per <id> *.md5 or *.asc are such metadata. 4) Do we need client-side metadata? 4.1) On repositories (how do we discover?) 4.2) A list of certain groups/ids needed for something (e.g. a project to build/run). 4.2 is broaching onto dependencies [or is dependencies], and I know different tools have different solutions/thoughts. I'd love to hear other folks views... Any other metadata I'm forgetting? regards, Adam -- Experience Sybase Technology... http://www.try.sybase.com
