> So it will be our problem if we would like
> to consider any constrains for version names (which IMHO are no realistic
> in short term (if ever) even inside ASF) .
> That's why I strongly believe that any discussion about artifact
versioning
> simply should not take place.
You don't want to slow down repository in order to attemtp to debate/define
some likely insoluble attribute. I respect that, and complete agree. Just
like other artifact metadata, we are agreeing to defer on them.
I am making the point that without understanding the version (to some
extent) we will not be able to have intelligent clients that can do anything
'intelligent' without metadata. We can't automatically clean up older copies
(we can't trust a timestamp), we can't selected "the best", we can't order,
etc.
We've agreed that we ought not inhibit apache teams/tools extending the
repository with metadata. A live and let live approach. All that I'm asking
is that we don't limit tools from benefits just because we don't wish to
define them.
> And we should assume that
> version = pchar*
Sorry, I don't have the definition of pchar to hand. Does it includes '-'?
Does it include period and the following '.jar'? I'm not so much trying to
define the version namespace, but allow us to parse the URIs without
ambiguity.
I think we need to take Tim's proposal as a good starting place and dissect
each part until we believe it to be tight. Just like Nick has started with
version location:
http://nagoya.apache.org/wiki/apachewiki.cgi?ASFRepository/ToDo
I feel we need to do this for group, for type, and so forth. I know it is
tedious, but (as I said before, sorry to repeat) this is probably *all* this
repository@ workgroup ought focus on/produce.
regards,
Adam
