[cgl_discussion] linux-ha as a PoC for CCM.1 Cluster Communication Service

Shureih, Tariq tariq.shureih at intel.com
Tue Jul 8 09:45:23 PDT 2003


I would have to somewhat agree with Patrick with the exclusion of the
"New kid on the block statement"; he has a point in that there seems to
be a lot of gap analysis on the part of PoC, specs and tech board
members but not a lot of input nor discussion with the
maintainers/owners of these projects.

Ok, maybe there is _SOME_ two way discussion with project maintainers
(IBM engineers is the example), however I have to say that more can and
should be done to involve the projects referred to in PoC and Specs and
pull them into the "Matrix" of CGL.

Julie (and Mika when you're back), let's discuss this a bit further and
maybe entertain the idea of establishing communication lines with these
Open Source projects we keep referencing.  It may even prove fruitful in
which they cooperate further with CGL and possibly expedite or branch
CGL specific development?

Patrick, you're a Kernel hacker ;-)  How about a CGL patch set like the
one Randy is maintaining for KJP which you push into LKML?  (holding my
breath)

--
Tariq Shureih
 

*Opinions are my own and don't reflect those of my employer*


> -----Original Message-----
> From: Patrick Mochel [mailto:mochel at osdl.org]
> Sent: Tuesday, July 08, 2003 9:23 AM
> To: Fleischer, Julie N
> Cc: Zhu, Yi; cgl_discussion at osdl.org
> Subject: RE: [cgl_discussion] linux-ha as a PoC for CCM.1 Cluster
> Communication Service
> 
> 
> On Tue, 8 Jul 2003, Fleischer, Julie N wrote:
> 
> > > Hi Julie,
> > >
> > > Given most of the features to NO, I don't think linux-ha is
> > > still suitable
> > > to be a PoC for the current requirement. Is it possible to
> > > change the CGL
> > > cluster communication requirement to be more generic? I mean we
don't
> > > define approach details in the spec. For example, we just say
> > > CGL shall
> > > provide a point-to-point and multipoint communication service
> > > but don't
> > > specify it must use a socket-based interface; we can say CGL
> > > shall provide
> > > a reliable communication service that detects communication
> > > failures but
> > > don't specify the communication mode must be connection or
> > > connectionless.
> > >
> > > Thanks,
> > > -yi
> >
> > I'll make a note to bring this up at the POC meeting.  As far as the
CGL
> Spec or POC tracking list, my understanding is that it's okay to leave
> projects, such as Linux-HA, that don't implement the full requirement
as
> references.  In the POC tracking list, the "gap analysis" column is
used
> to define which projects will/won't implement the full requirement.
> >
> > I'll bring up what you suggest:  "Do we have a viable option in the
CGL
> 2.0 timeframe for the requirement as it is worded?  What about if the
> requirement were more generic?"
> 
> Something that has been occuring to me in recent flurry of project
> analysis to fulfill CGL requirements is that the projects and/or
people
> involved are not contacted at all to determine whether or not they may
or
> may not meet your requirements in the future.
> 
> You're simply speculating on the viability of these projects to meet
your
> needs, unless they explicitly state that they fulfill the
requirements. As
> was already seen with the 'Persistant Device Naming' (or whatever it
is
> called this week) Requirement, the collective speculation overlooked
the
> potential of the udev (and related) projects.
> 
> I fear that other speculation may lead to overlooking of other
projects,
> which will potentially lead to someone wasting a lot of time
reinventing
> something rather than reusing and improving something else.
> 
> The current practice is not very community-oriented.  Although the CGL
> project has not had a great reception in the past, it is imperative
that
> you communicate with the other Open Source projects you're evaluating
> using.
> 
> You guys are (still) the new kids at school. No one is going to come
> running over to be your friends; you have to talk to the other kids
and
> make yourselves likeable. :)
> 
> 
> 	-pat
> 
> _______________________________________________
> cgl_discussion mailing list
> cgl_discussion at lists.osdl.org
> http://lists.osdl.org/mailman/listinfo/cgl_discussion




More information about the cgl_discussion mailing list