> |------------------------------------------------------------| > | get/put to cache with group/name/blob-version semantics, | > | security on transmission, integrity checking, | > | and reliability of service | > |------------------------------------------------------------| > > ... and the starting point is the bottom layer.
I can agree with starting at the bottom. I think this (above) is about it. I would like to offer both opaque(blob) version and structured/parsed, but I do (now) realize that folks need the ability to have the former (as a starting point). I've read the Avalon Repository code at: http://svn.apache.org/repos/asf/avalon/trunk/runtime/repository and I'm very pleased to see the significant similarities. I like the clarity/simplicity that Avalon Repository has, but I also like some of the features that Depot provides (e.g. ability to use java.net.URL or HttpClient or VFS as environment allows). I do like that we have Filters/Selectors in common (Depot has Comparators also, 'cos we think order is possible). I could go on, but not here. Again, I see more in common than I see in divergence. I'd be tempted to suggest we import the Repository code into Depot (as a separate ClassLoader project) and slowly (or quickly) migrate as much into the Updater project, or use Updater concepts. I think there are interesting choices to be made, and we could work that merge (over a suitable time) with some Wiki documentation and/or scheduled group chats. [For example, the Artifact and (Depot) Resource classes are next to identical, we need to document what we want, and merge them.] *If* we can all open our minds (and suppress our egos) sufficiently to allow this form of merge (without tonnes of blah/blah back and forth on minutiae) then I think we'd have one heck of a lively/healthy development team (community), worthy [IMHO] of Apache TLP. regards, Adam