On Sat, 14 Jun 2008 15:25:53 +0200
Luca Barbato <lu_zero@g.o> wrote:
> Ciaran McCreesh wrote:
> > * What generation looks like.
> 
> Mostly implementation detail? Somebody seems to have ideas there and
> I like to heard ideas from others as well.

It's not a detail. It's extremely important, and we can't discuss this
sanely until we have detailed proposals for this.

> > * How to select which ebuilds to trigger generation for.
> 
> I'm fond of sets and I'd extend maint to be feeded to sets other than 
> world and all.

Again, details.

> > * When specifically to trigger generation.
> 
> User decision or triggered by sync depending on a FEATURE.

Again, details. This one *really* needs to be worked out, since it's
getting in the way of discussing other implications.

> > * The security implications of the previous point.
> 
> None?

Well, generated revision numbers may have to be stored somewhere.
Should a normal user be able to force a generation? If not, this means
having to generate templates at sync time for every scm package in the
tree, which in turn means scanning every ebuild in the tree.

> > * What the impact upon upstream servers is, if any.
> 
> Shared with glep 54

GLEP 54 doesn't use upstream scm revision information at all. Are you
definitely saying that your proposal will not be tied in to upstream
revisions in any way?

-- 
Ciaran McCreesh