| Re: [equinox-dev] Support for RFC 119 |
Tom
Thomas Watson wrote:
> > Scott Lewis wrote:
<stuff deleted>
> > Thanks much. One question: If we/ECF is going to have or create the
> > bundle that exports the distribution provider and discovery interfaces
> > from the r4.2 cmpn...should this bundle be in the equinox namespace (and
> > if so what?) or should it be in the ecf namespace (i.e. in
> > org.eclipse.ecf core bundle or some such)?
>
> My preference is the package is exported (and imported to allow for
> substitutability) by the bundle implementing RFC 119. I don't think
> it should be exported by the ecf core bundle unless that is the
> bundle that is actually going to implement RFC 119.
>
> One of the Equinox missions is to be the place where OSGi implementations
> are done within Eclipse. We need to decide if that means it is the only
> place at eclipse where this is done. If so then the namespace would become
> something like org.eclipse.equinox.distributed (seems like an awfully
> general name though).
Jeff and I discussed this. We do not think we can/should force all OSGi
specification implementations to reside in Equinox. In this case, if
RFC 119 is in the same general realm as ECF then it makes sense for the
ECF project to provide an implementation of RFC 119.
It is perfectly fine for a bundle in the org.eclipse.ecf namespace to
export an OSGi specification package and provide an implementation of
it.
>
> The other thing to consider is that the Enterprise specification within
> OSGi has a later schedule than the core specification. Currently for
> Galileo Equinox is only implementing specifications coming out of the
> Core Platform Expert Group (CPEG) which has a release schedule that
> aligns nicely with the Galileo release schedule.
>
> We may need to consider incubating this effort (RFC 119 implementation)
> for the Galileo release and releasing a graduated version in a later
> release. Potentially we could graduate in a point release of
> Galileo. Another option is to make the OSGi API provisional (this
> means marking the exports as x-internal) until the Enterprise
> Export Group finalizes the specification.
>
I did not see your opinion on this issue Scott. By the time Galileo API
freeze occurs we need to have confidence that the RFC 119 API will be
final. I don't think that we can make that assumption. We may need
to take the same approach that is being taken for Composite Bundles
in Equinox. The API will probably need to be considered provisional API
(x-internal) until the specification becomes final (sometime in June).
That puts us into the first point release of Galileo.
Tom.