Responses in line.
Wayne
On 06/12/2012 12:42 PM, Marcel Bruch wrote:
I'm picking up Wayne's and Konstantin's ideas:
How would the process of adding a new project to an project
incubator look like? Paperwork for one or two committers, a
project description with a mission statement, initial code IP
check, "and go!"? One or more project committers could take
over the mentor part and help with things like IP, CQs, and
processes etc.?
That sounds right. Again, however, the scope of the incubator needs
to fall inside the scope of the parent project.
http://www.eclipse.org/projects/dev_process/development_process_2011.php#4_9_Incubators
Note that a creation review is still required, but no proposal. I'm
completely happy with a Bugzilla record being the review document.
FWIW, if you just include a slide in the 1.0 release documentation,
we can create an incubator based on that.
The incubator is subject to the same rules and process regarding
committers. The PMC does allow more latitude with regard to merit.
Even with that latitude, however, we would still take issue with
making somebody a committer "just because"; we generally expect
committers to come with code (and code to come with committers).
For Recommenders, I would like to take the opportunity to
attract other research groups and set up the infrastructure and
guidelines to make the move to an Eclipse Incubator attractive.
I understand this as an experiment and an offering we could
present to interested research groups and see how the responses
are (before taking any actions upfront).
Would technology-pmc and foundation agree with this? If so, I
would announce this to "Recommendation Systems in Software
Engineering" (RSSE) and "Mining Software Repositories" (MSR)
research communities in, say, July to get their feedback on
this.
+1
Make sure that they are aware a that there is a process for
graduating code out of an incubator. Note also, that the EDP
explicitly states that incubators do not have releases. If they want
to "release" something, that's a pretty good indicator that it's
time to create a real project.
Marcel
On 12.06.2012, at 17:39, Wayne Beaton wrote:
We can pretty easily create incubators under an
existing project. With graduation, for example, Code
Recommenders can request that an incubator be created
for related work. The incubator does have to work within
the scope of the parent project.
I'm keen to make the process of creating a project
easier. I do not, however, have any designs on making it
as easy as creating a new project at Source Forge or
GitHub. That's just not the space we live in.
I'm not convinced that every "research" project should
be created under Technology. Technology was originally
created as an incubator for new ideas that either grew
and moved to Tools or some other appropriate home; or
reached some natural conclusion and were terminated.
Over time, Technology has become a place for stuff that
doesn't fit anywhere else.
There's really no reason why a "research" project
couldn't be started under Mylyn, for example, where it
would receive far more attention from the ALM community
(assuming that's the community being targeted).
What is the point of bringing a "research" project to
Eclipse? I find it hard to believe that the main
motivation would be to connect with other researchers.
Rather, I'd expect that it would be to extend the reach
into the broader community and eco-system.
This is is an area where I think we could do better.
I've been trying to help new projects sort out the most
appropriate home and otherwise set themselves up for
success. But this has the effect of prolonging the
proposal process and making it even harder to get
started (in a good way, though).
FWIW, Mylyn is the ultimate research success story. It
did start off life as a research project in Technology
and moved to Tools before being promoted to a Top-Level
project. Maybe Mik has a different opinion, but I don't
think that being classified as "research" is what lead
to success. Success, I believe, came from a combination
of game-changing technology, and hard work. Lots of hard
work. This is the same combination that we're seeing in
Code Recommenders.
Wayne
On 06/11/2012 12:26 PM, Marcel Bruch wrote:
Resend from different address. + added
comment on CBI.
On 11.06.2012, at 18:08, Marcel Bruch wrote:
Hi Ian,
Hi Wayne,
IMHO research projects are most interested
in more publicity and assistance in building a
community. I don't think that a forge or a
wiki page helps much here. Even forums won't
help as developers don't know anything about
the projects behind them.
As you both propose, a page that lists all
ongoing research projects under, say, http://eclipse.org/research/projects
would be nice. Such a page could be driven by
some tags in the Eclipse Marketplace with
relatively low effort I think.
Wayne, regarding the proposal process. I
know it's there for good reasons. But at the
same time I think that projects like Snipmatch
wouldn't have considered joining Eclipse if
they had to declare a committer team, a
project proposal, a project plan and the like.
This is too heavyweight.
I think Incubators are the right way to go.
They don't need a project proposal nor naming
a committer team nor declaring a project plan
(even it would be great to have one for
Snipmatch ;)) I think, there should be a
lightweight process that enables research
projects to join another project and naming
one or two committers without the need of a
big commit history. Then, there should be
incubator update sites that make is easy that
projects get their tools out to the users. The
hosting project should also make the marketing
(blog-posts, tweets etc.) to get these tools
out to the developers - at least enable
incubators to use existing channels. CBI is
good as it enables quality assurance and
build automation with minimal effort.
And these things and expectations should be
documented somewhere. I had hard times to
figure out where to ask for permission of
whatever. Something like the committer
resources wiki page for research projects
would be nice.
Wayne, I think technology project is fine
if one of it's goals is to host research
projects. It should just be more present,
i.e., more actively advertising itself as
such. FWIW, I wasn't sure which top-level to
pick as tools, mylyn and technology all looked
good to me. I'm not sure if another top-level
project like "research" would help much. Maybe
if eclipse would make it a large sandbox for
research projects as I described above :)
These are just some unfiltered thoughts I
had after discussing ideas with some
researchers.
Thanks,
Marcel
On 11.06.2012, at 15:59, Wayne Beaton
wrote:
Hi Marcel. I'm a member of the
Technology PMC; of course I'm listening
(FWIW, I listen on all PMC list and a
great many project lists).
We've been doing a lot of work lately to
try and make a few things easier for
projects. The Common Build
Infrastructure should make building a
lot easier. A lot of projects just use
the metadata driven websites rather than
create their own; we're doing some work
to make this even better and easier with
the new Project Management
Infrastructure initiative.
Can you be more specific about what
parts of the entry barrier should be
lowered? Is the proposal process too
difficult/too time-consuming?
The Technology project was originally
intended (at least partially) as a place
for research projects. I think it's fair
to say that it has evolved away from
that. Maybe, as Ian suggests, we can
start by making university research
projects be more prominent on the site.
We can do all of the things that you
suggest within the scope of the
Technology Project. Or maybe it's time
to create a new top-level project.
Wayne
On 11.06.2012, at 15:19, Ian
Skerrett wrote:
Marcel,
I think we should always be looking to
improve how we reach out to different
communities, the research community
certainly being an important one.
EclipseLabs was an attempt to create
an extend community for projects that
didn't want to be 'official' projects
but wanted to be closer to the
community. It was setup so the
researchers didn't have to worry about
setting up their own forge and the
project code was available in the
open.
It seems like you are looking for a
bit more exposure for research
projects or 'home' for these types of
projects. I wonder if some type of
wiki page and/or forum would be a
starting point?
Ian
On 06/10/2012 05:10 AM, Marcel Bruch
wrote:
Hi technology-pmc,
If this reads a bit like a rant,
please excuse. It's not. Its intent
is to get one or two ideas how to
improve the current situation.
I'm just back from a research
conference and have been asked by a
bunch of researchers how potential
collaborations with Eclipse and Code
Recommenders in particular could
look like. The scope of these works
varies from applying Natural
Language processing (NLP) on
documentation, bringing NLP into
code completion, integrating Code
Recommenders into Code Bubbles,
developing a parameter guessing
recommender, collaborations on code
search engines, mining on user
interactions, and generally
extending the idea of IDE 2.0 for
lots of other ideas.
I don't think that many of these
ideas will actually turn into code
at eclipse.org
but if a few projects or ideas will
do so, it would be a great success.
I wonder whether Eclipse could do
more to get more research ideas into
Eclipse and provide them a platform
for their work. In my opinion
putting something into the
marketplace is not enough - research
people don't get the feeling that
they have a huge outreach there.
Can't we do a little more that they
get the feeling of being "part of
Eclipse" rather than "yet another
research prototype using Eclipse"?
Or can we lower the entrance
barrier for research at Eclipse? I
know that eclipselabs.org
was (also) designed for this case,
but do they work as expected? And:
is providing a repository a useful
support? What distinguishes it from
SourceForge or GitHub? I think these
research projects should be coupled
more to existing Eclipse projects;
they should be treated more like
incubators with associated
(top-level?) projects giving them a
platform for instance with an
aggregator update site, blog posts
etc.
Also, projects still have to
provide the whole infrastructure
like a build server, a web server
etc. on their own. We, for instance,
have a shadow infrastructure with
Bugzilla, Gerrit, Jenkins etc.
running at the university from the
first days which was a huge invest
we had to make upfront. And at the
end everything still stays in the
university network. This doesn't
feel like open source then and such
a huge support from my (very
personal) viewpoint.
A few thoughts on whether or how
we can change some things a little
would be great. I hope the
technology-pmc list is appropriate
for this as I'm only hoping for some
small changes inside technology
top-level project but not for
changes in the Eclipse bylaws ;) But
maybe this should just go to the
foundation. If so, I hope Wayne is
listening.
Thanks,
Marcel
P.S.: I know the
discussions about
"researchers want to
publish papers and don't
want to support tools
for long time". This is
not the direction I
would like to take in
this post. It's about
simplifying the process
iff someone wants to go
a few steps further -
like we did with
recommenders. It just
doesn't need to be that
hard as it was for us.
_______________________________________________
recommenders-dev mailing list
recommenders-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/recommenders-dev
--
Wayne Beaton
The Eclipse Foundation
Twitter: @waynebeaton
Explore Eclipse
Projects
recommenders-dev mailing list
recommenders-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/recommenders-dev
_______________________________________________
recommenders-dev mailing list
recommenders-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/recommenders-dev
_______________________________________________
technology-pmc mailing list
technology-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/technology-pmc
--
Wayne Beaton
The Eclipse Foundation
Twitter: @waynebeaton
Explore Eclipse
Projects
technology-pmc mailing list
technology-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/technology-pmc
_______________________________________________
technology-pmc mailing list
technology-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/technology-pmc
--
Wayne Beaton
The Eclipse Foundation
Twitter: @waynebeaton
Explore Eclipse
Projects |
