Re: Adding extra metadata to Yocto/Poky

[Richard Purdie <rpurdie@linux.intel.com>]

On Wed, 2010-12-08 at 07:08 -0700, Chris Larson wrote:
> On Wed, Dec 8, 2010 at 5:49 AM, Richard Purdie <rpurdie@linux.intel.com> wrote:
> > We're in a position where we have people wanting to extend Poky but
> > wondering exactly how to add their own layer and make Poky useful for
> > their specific use cases.
> >
> > This is a topic I want to work with the OE community on and have
> > discussion about but equally, people are interested at the moment and we
> > need something people can experiment with now so we can see what works
> > and what doesn't.
> >
> > With this in mind I'm proposing we create a "poky-extras" repository
> > which contains multiple user contributed layers.
> >
> > My proposed ground rules are:
> >
> > * it has the OE style open contributions model
> > * consists of a set of layers, each clearly identified at the top level
> > * each layer has a clearly named maintainer or maintainers
> > * people respect the layer maintainers
> > * layers are split into logical "topics" where at all possible
> >
> > Ultimately, the repo structure and some of the details could change but
> > I'd like to experiment and see what we can come up with which is
> > probably the only way we're going to move forward.
> >
> > Any comments, objections or better ideas? :)
> 
> This seems fine to me as an area to hash out good practices and
> experiment, as long as we don't expect to use this structure long
> term.  It makes sense for the core layers to be in one repo, but it
> makes more sense to use one git repository per external layer, imo.
> Is that the intent, or am I misunderstanding its purpose?

That is the intent but we don't currently have the tooling to make repos
per external layer work nicely for the end user (not that I think its
too hard to create). This repo give us a way to experiment with
splitting things up whilst still not hurting the end user experience too
much.

So its not the final solution, just what I think might be a good way to
start hashing things out.

Cheers,

Richard