Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [soa-iwg] ECF remote services

Zsolt Beothy-Elo wrote:
Hi Scott,
congratulations to you and the ECF team for the promptly implemantion and release of the spec. What I am still missing is the tooling aspect

WRT tooling, four points:

1) A large part of the whole *point* of the OSGi remote services standardization is to allow the use/reuse of *local* OSGi services tooling for remote services. In the case of Eclipse, such tooling is already well established and heavily used via PDE, declarative services, extension registry tooling, etc. All of this tooling is immediately usable with ECF's remote services implementation...precisely because the OSGi standardization supports such a model.

2) ECF has UI/tooling for the most difficult (and communications-specific) part of the use/debugging of remote services...and that's the dynamic network service discovery. Right now, we provide an extensible discovered services browser, so that the use of discovery protocols can be verified to be working when debugging/testing.

3) ECF is a runtime project, and in keeping with that and the fact that we have plenty that we've committed to do, we don't intend to focus on tooling. Having said this, however, I (Scott) am also in the process of implementing a SOAP-based remote services provider, which can/will/could use tooling produced in other project (e.g. WTP-tooling, possibly Swordfish tooling, etc). Further, we are always open to community contributions. In general, since we are a runtime project, and there are a lot of excellent tools projects at EF, we will (intentionally) concentrate on implementing the OSGi runtime standards, and work collaboratively with other projects and the broader community to deliver tooling.

4) Finally, I think it's inappropriate to expect a runtime project to necessarily also produce tooling...especially when there are so many tools projects at EF. For example, consider Equinox...yes they do work with the platform and PDE teams on tooling (as we do), but Equinox itself is not (and I would argue shouldn't be), expected to implement tooling themselves.

and also the integration with the other components.

What integration are you talking about? We have the following integration already

1) Integration with OSGi mechanisms (service registry) as dictated by the OSGi spec
2) Integration with Equinox (implementation of the OSGi framework)
3) Integration with Eclipse PDE, declarative services tooling...and other Eclipse-based tooling for constructing, using, debugging OSGi services.

For an implementation of an OSGi standard like remote services...on EF's implementation of that standard (Equinox), what other integrations are you suggesting is appropriate/necessary?

I do not expect them to be available today, but we should have a roadmap which addresses these topics and commitments to implement the roadmap.

As for tooling...like I said...we are leaving the tooling primarily to the relevant *tools projects*. For the areas of tooling that are unique (like discovery) we have (and will continue) to do some further things there...and in general work collaboratively with the other projects focusing on tooling (e.g. EMF, PDE, etc).

If you want our roadmap, please see the ECF plan: http://www.eclipse.org/projects/project-plan.php?projectid=rt.ecf and/or the discussion on the ecf-dev mailing list. And if you want to add to the roadmap, then you are encouraged to contribute the resources to implement the desired additions.

In a second step we should then discuss what has to be in a first contribution for the EPP package to make it useful.

My assertion is that it should include:

The OSGi remote services implementation bundles

org.eclipse.ecf.osgi.services.discovery
org.eclipse.ecf.osgi.services.distribution

and dependencies...which are:

org.eclipse.ecf.remoteservice (ECF remote services API)
org.eclipse.ecf.discovery (ECF discovery API)

and their dependencies (the ECF core bundles....which are already distributed with Eclipse since they are part of p2...but would have to be included in a non-eclipse/runtime).

Further, it should probably include at least 2 provider implementations, so that things can/could be used immediately and 'out of the box'...e.g. the r-OSGi provider, and the ECF generic provider (along with their dependencies).

That makes a total of about 10 bundles, and < 1M of jars.

Scott



Back to the top