Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [] A suggested topic for PlanningCouncil Discussion

Hi Ed,

Ed Merks wrote:

This is an excellent summary of the issue that lies beneath this week's
confusion of ideas.  From the perspective you've outlined, the release
train simply delivers the raw materials and a commercial vendor would
refine these to assemble their products.   Eclipse's packaging efforts are
the open source equivalent of such a refinement process.  In both cases,
having a train well stocked with a wealth of commodities provides the raw
materials to drive the effort.  Of course a carload of steel will be of
more value than a carload of iron ore, so there will be and should be
pressure on the level of refinement in the train itself,  but we have to
recognize that the train itself might not deliver a fully manufactured car
and that perhaps this should not even be its purpose.  After all, high
quality steel and component parts is what the car manufactures need to
assemble their cars, so to me that's what the focus of the train should be.

To abuse this metaphor a little:  why wouldn't/shouldn't the train be focused on the needs of the consumers (car buyers) as much as or more than the car manufacturers? 

I know and appreciate that the car manufacturers have a certain set of needs here, and that those are important to the community.  I'm not trying to minimize those needs.  But it seems to me that there are other important community needs here too...e.g. consumer needs to have new stuff as part of Eclipse (innovation), simple end-user install/upgrade, all projects need exposure/distribution in order to succeed...particularly relatively new areas (e.g. RCP), and others.


Certainly from a modeling perspective, Rich and I are very keen on
producing a Modeling package from the Ganymede component parts.  This is
badly needed by those in modeling community who are currently served a
confusing bucket of nuts and bolts; they've told us this a great many
times.    It's important to note that many of these nuts and bolts are
incubating projects, so I would be extremely unhappy if those projects are
not allowed on the Ganymede train.  I will continue to question efforts in
that direction and I will feel compelled to remind people the extent to
which the Ganymede train relies on locomotives like EMF to make it work.

I'll be more than happy to follow the collectively agreed to must do's for
Europa, but if those must do's start to take the form of exclusionary
tactics, I'll reconsider whether the goodwill demonstrated by following the
collective rules is being rewarded in a way that benefits the users who
want to see an integrated Modeling package.   I do expect to set a high bar
for what goes in the Modeling package and I expect to test it well.  I also
expect folks who want to ship in the Modeling package to help define those
rules and most importantly to follow them.  I don't see the Planning
Council's place right now to be setting those rules for all of us at the
level of the train.  I can imagine over the years that the experience from
these packaging efforts will feed back into rules that produce more refined
results for the train over all, but I think it's premature to do that now.

I'll balance my comments above by saying that I can also see value in
taking all of Ganymede and producing one giant package or equivalently
sucking in the entire population of features from the update manager, not
for the purpose of redistribution, but for the purpose of testing.  The
question is, who will do that testing?  Will there be a community
interested in this?  I'm quite sure that such a result does not have value
to the Modeling community, nor likely to any specific community, so I'm not
sure my efforts toward producing a Modeling package will be well served by
testing the complete mishmash.  If others will step up, I will consider
doing my part.   In the end, while I understand the desire to reduce the
mishmash train down to something we might call a product, I'm not convinced
that effort isn't somewhat misguided or misplaced.

I'm sure another round of debate will involve what packages are "special"
and hence which packages will get to be on the main download page.  My
sense right now is that we have four flavours of vanilla: plain, classic,
French, and natural honey.   Of course some will argue that having one
hundred and one flavors will make choosing too difficult.  After all, it's
already difficult to choose from the four flavors of vanilla.  So even the
solution of providing integrated packages also leads to new problems...

Ed Merks/Toronto/IBM@IBMCA
mailto: merks@xxxxxxxxxx
905-413-3265  (t/l 313)

             /IBM@IBMCA                                                 To 
             Sent by:         
             ing-council-bounc                                          cc 
                                       Re: []  
             11/03/2007 07:06          A suggested topic     for           
             PM                        PlanningCouncil Discussion          
             Please respond to                                             

To me, the hard problem is *how* do you create a unified whole out of a
mish-mash of projects?  It's certainly easier to do with fewer projects
than the number currently planning to go into Ganymede.  But are we
honestly going to be able to come up with a good set of criteria for
inclusion, much less a way to fairly apply that criteria?  And can we
possibly achieve that without creating major ill will?

If we wanted to follow the example of Linux distributions, they definitely
don't do it that way.  A distribution is produced by a whole other set of
people that define their criteria and do the packaging, integration
testing, upstream bug reporting and even, where necessary, fix problems
locally.  They get to make the decisions because they do the legwork.  And,
of course, it behooves the projects to help them out because they're the
channel to the users.

I would think that the parallel would be to use the output of Ganymede as
input into an integration/packaging effort.  Are there any interests in the
Eclipse community who would step up and do this kind of work?


Dave Steinberg
Rational Software - IBM Toronto Lab

Nick Boldt wrote on 11/02/2007 11:31:07 PM:

So, then, you're for the idea of merit-based tiers? Something akin
to package repos like main, restricted, universe, multiverse,
testing, stable, etc.

Seems entirely reasonable, IMHO, especially if we don't call them
"Hi Fi" and "Simultaneous" [1], but rather "main" and "multiverse".


Nick _______________________________________________ mailing list
_______________________________________________ mailing list

_______________________________________________ mailing list

Back to the top