| Re: [equinox-dev] Equinox Subprojects |
|
The different RT projects actually operate as separate projects.
Virgo, for example, has its builds, downloads, website, etc. The Equinox subprojects, however, do not operate as separate projects. They operate as ACLs. p2 is a bit of an exception, but still leverages much from the parent project (which is fine). I would like to kill off the inactive project in the short term it appears to provide no value. I would like to kill off the "website" project as it's existence is an ugly hack that is a bogus project that skews our numbers. However, there's no particular urgency to do this. I'm keen to help sort out a reasonable alternative. I'm sceptical of the value of the other subprojects (besides p2 and the incubator). But I leave it the project and PMC discretion to decide what to do. There is no urgency on my part to split p2 from Equinox. Since p2 depends on Equinox for builds and downloads, the split does represent some work. I leave it to the parties involved to decide if it is worth the investment. Wayne On 12/06/2012 03:37 PM, John Arthorne wrote: By this kind of argument one could argue all of RT should only have one committer group and we use social conventions to avoid writing in the wrong area. Personally I see p2 and Equinox framework as quite different projects. Equinox framework is agnostic of what provisioning technology is used to install it, and it has no knowledge or dependency on p2. Similarly p2 could be used to provision other OSGi implementations of even non-OSGi applications. The two projects have no overlap in active committers that I can think of. They share a website repository but have separate web sites. I really don't see any violation of the spirit of the EDP in having them as separate projects. Just my $0.02... John From: Wayne Beaton <wayne@xxxxxxxxxxx> To: equinox-dev@xxxxxxxxxxx, Date: 12/06/2012 01:02 PM Subject: Re: [equinox-dev] Equinox Subprojects Sent by: equinox-dev-bounces@xxxxxxxxxxx Is there real need to protect different functional areas within the project? Is there danger that a "website" committer might start hacking p2 code? I'm having trouble understanding why subprojects are required at all. Equinox builds and releases as a single thing. There is a single website. All of the subproject repositories are in a single location. The only reason that I can think of to have subprojects is so that you can have ACLs against different functional areas within the same project. While this may align with the wording of the EDP, I think that it violates the spirit of it. Is there any reason why we can't just collapse *all* of the Equinox subprojects into the parent and be be left with rt.equinox? The CDT project, for example, does this. They have one set of committers. Within the project they manage who accesses what by social convention rather than enforced access control. Is there valid concern that we need to keep p2 committers from touching framework stuff? Collapsing would solve the website problem, wouldn't it? Wayne On 12/06/2012 10:00 AM, John Arthorne wrote: +1 on terminating rt.equinox.security. Effectively we did this during the Git migration and forgot to do the full process. I don't care either way about combining bundles+framework. They do have fairly distinct committer lists and they seem to be functioning fine in their current form. On the other hand I would trust all the committers to only work in their area of expertise so I have nothing against combining them. I tend to agree with Wayne that effectively they operate as a single project. On the website, the other big issue is that the Equinox download directory is owned by the website project. So between the downloads and web sites I do think all the Equinox committers need to have commit rights. Whether that is done via a separate project or ACL magic from the webmasters like Platform does, I don't mind either way. When dealing with several different directories I tend to think the ACL approach would be a pain to manage. Looking at the members of the website ACL, it does look in need of some cleanup at least - even McQ is a committer there :) John From: Thomas Watson <tjwatson@xxxxxxxxxx> To: Equinox development mailing list <equinox-dev@xxxxxxxxxxx>, Date: 12/06/2012 08:38 AM Subject: Re: [equinox-dev] Equinox Subprojects Sent by: equinox-dev-bounces@xxxxxxxxxxx rt.equinox.security should have been roled up into rt.equinox.bundles. At least we moved all the security code into the rt.equinox.bundles repository and have one commit group for that repository. So technically it is a candidate for termination, but the code did not go away. Personally I would be fine with combining rt.equinox.bundles and rt.equinox.framework into one project under rt.equinox and leaving rt.equinox.p2 as the sole subproject. This means all rt.equinox.bundles committers would gain commit rights to the rt.equinox.framework repo and vise-versa. I'm not sure what to do about rt.equinox.website project. If you have an easy way to also combine it into rt.equinox then that is fine. But we must allow rt.equinox.p2 committers to still have access to the web site repository. Others have opinions? Tom Wayne Beaton ---12/05/2012 04:29:34 PM---I just noticed that some of the Equinox subprojects do not have any source repositories listed [1]. From: Wayne Beaton <wayne@xxxxxxxxxxx> To: equinox-dev@xxxxxxxxxxx, Date: 12/05/2012 04:29 PM Subject: [equinox-dev] Equinox Subprojects Sent by: equinox-dev-bounces@xxxxxxxxxxx I just noticed that some of the Equinox subprojects do not have any source repositories listed [1]. In fact, most of them have no metadata specified at all. Further, upon inspection, it appears that the rt.equinox.security project has no resources associated with it (no Git repository, no downloads that I can detect, no website). Is this project still viable, or is it a candidate for termination? The rt.equinox.website project is a hold over from the bad-old-days. Is that project still required? Can we kill it and assign the website repository to rt.equinox? Is it still valuable to have Equinox subprojects at all? Based on the use of the projects, it seems that the only purpose is to keep the committer lists distinct. Is this still necessary? AFAICT, only rt.equinox.p2 seems to be operating as a separate project. Does it make sense to consider rolling the rest of the projects up into the parent? Thanks, Wayne [1] http://eclipse.org/projects/tools/status.php?cvs=0&git=0&svn=0&top=rt_______________________________________________ equinox-dev mailing list equinox-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/equinox-dev |
