Coda File System

Re: Coda status quo and future

From: M. Satyanarayanan <satya_at_cs.cmu.edu>
Date: Fri, 23 Jul 2010 14:26:50 -0400
Dear Jaap,
  Thank you for the compliment --- "mouth-watering" is a wonderful term!
As the prime mover of the Coda project (i.e., "chief cook and bottle washer"
or "chief sanitation engineer", depending on your perspective :-))
let me try to give you the "big picture" answer to your question.

  Coda works well enough that a small group of us at CMU have all our files
in it (personal and project) and rely on it very deeply for all our work every day.
So it is definitely very real in that sense.   However,  we also know of many things
that we could do to make the implementation much better (including reimplementation
of parts).   However that takes sustained time and effort of high quality people.   This can
happen in one of two ways: (a) volunteers like yourself do the work or (b) I receive
substantial enough and stable enough funding for my team specifically for this
engineering effort.

  Coda today is in that gray zone between "leading edge research"  and "robust product".
My research group sustains the current code base and improves it in modest ways,
but unless something major happens to change things,  we simply don't have the
resources to actually make all the improvements that we know can be made.    We will
continue to sustain what we have, and will improve it in modest ways   using the
small amounts of people time we can spare from the other projects for which we have
research funding.   In other words, we have the will and the knowledge but not
the resources.  Funding agencies like DARPA and NSF that provided
the original funding for Coda many years ago consider it "done" --- i.e. from a pure
research perspective it has produced valuable insights and publications. 
So getting resources from them to carry the work forward is not feasible.

  In principle, this is exactly the problem solved by the open source development
model.  Lots of people can give small amounts of time each, and the total impact can
be large.    However, although Coda is open source, we have not seen this happen
(yet).    We would be delighted to have such an effort be catalyzed, but perhaps it
requires organizational, leadership or other skills that we simply lack.   If you and others
on the mailing list possess these skills, and can help catalyze a substantial volunteer effort
that would be fantastic.   I would be happy to host a meeting at Carnegie Mellon
to kick off such an effort, if there is enough interest.  A day or two of deep
immersion may be needed  to share in-depth knowledge of the code base and to
develop a work plan.

  Hope this helps explain what you see.  We would LOVE to see a major Coda 
re-engineering effort, so if you and others on this mailing list would like to 
see that happen please speak up, and consider playing a leadership role in the effort.
Or, convince someone to write us a large check :-)
Cheers
                  --- Satya

p.s .  I'll be off email (on vacation) for about 10 days starting tomorrow, but will
        catch up when I return.   Jan and others in my group can contribute to the
        discussion in my absence.   But it is the rest of the community on this mailing list
        that really is the key to moving forward.




On Friday 23 July 2010 13:15:59 Jaap Winius wrote:
> Hi all,
> 
> To me, the Coda file system seems like an amazing concept. I'm  
> currently studying OpenAFS, but Coda has all kinds of mouth-watering  
> featured that OpenAFS doesn't, such as disconnected mode, intrinsic  
> replication of RW volumes, redundancy across server, etc.
> 
> Unfortunately, rumor has it that Coda never was intended as and is  
> never going to be suitable as a production file system -- that it's  
> really just a research project. But, perhaps those people are not  
> aware that there have been quite a few releases during the past  
> decade, with the last one only 18 months old. However, development  
> does seem to have tapered off.
> 
> So, how should the current release be viewed? Can it be recommend for anything
> beyond experimentation? And what does the future of the project look  
> like? Does it actually have one? Will it ever be considered as a  
> serious alternative to OpenAFS (for optimists)? Should we keen an eye  
> on it and hold our breath?
> 
> Thanks,
> 
> Jaap
> 
> 
> 
Received on 2010-07-23 14:27:06