Shawn Walker wrote:
Brock Pytlik wrote:
Shawn Walker wrote:
Shawn Walker wrote:
Tom Mueller (plain-text) wrote:
[snip]

If this is a correct understanding, would it be also be correct that with this proposal, bug 7653 need to be changed to talk about streams rather than repositories?

No, I suspect we'll still have separate repositories for software delivered by different groups, but each of those repositories may have multiple streams. For example, the '/contrib' and '/webstack' repositories need to stay separate, but could possibly offer their own streams.

So one obvious hole here is that we haven't yet figured out what to do with software offerings that don't really fit the 'stream' concept. For example, the stream concept works well with a set of packages that all move together (e.g. dev, release). But, it doesn't work so well for packages that merely exist together (e.g. contrib, pending).
One possibility is that pending and contrib should really be streams as well (under a more general "community" publisher or something like that).

Yes, though I suspect that for process and simplicity, that it is likely that the providers of them would keep them as separate repositories.

Perhaps. It's something we should discuss with the providers if we decide to go this direction.
The thought so far has been that if a publisher defines a stream, and a corresponding incorporation package exists that is tagged with that stream, then that is used to filter package listings and also used to move between streams. (e.g. if 'dev' 'entire' is installed, then move to 'release' entire, and follow normal package constraint application).

We could solve the cases such as contrib and pending by saying that if a corresponding, tagged incorporation doesn't exist for a defined stream, that the user can't select it or anything else, and it simply exists for definition. I believe we're still looking for suggestions about how to deal with that particular aspect.
You're suggesting that there isn't a one-to-one mapping between streams and stream-tagged incorporations?

In the case of contrib and pending, correct; there is no overall incorporation that restricts the versions of the packages contained within.

I'd suggest a different approach. The streams have incorporations, but don't define them down to the version level, merely the packages. That would mean that no special casing would be needed for those types of streams, and would allow pending and contrib to be declared incompatible with each other (which I think is what we'd want to do) using the standard method instead of more unique solutions.

Brock
Cheers,

_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to