[cddlm] FW: CDDLM-WG completion criteria?

[Toft, Peter]

Hi,

I just want to forward to the group the following very helpful response
from Stephen Pickles, regarding a question I asked about the completion
criteria for CDDLM.

Regards,
Peter

-----Original Message-----
From: Stephen M Pickles [mailto:Stephen.Pickles at manchester.ac.uk] 
Sent: 25 September 2006 19:10
To: Toft, Peter; Hiro Kishimoto
Subject: RE: CDDLM-WG completion criteria?

Hi Peter, and welcome.

A group is finished once it has completed its charter.
(You should check that the charter still matches the
group's plans). For CDDLM, I imagine this means
bringing all of the specifications you list through
to "Recommendation" status - that's when they become
OGF standards. 

You're close to having "Proposed Recommendations"
for these specs. A proposed recommendation is one
that's been through public comment and is published
in the GFD document series as a proposed recommendation.
That is, it's listed on
  http://www.ogf.org/gf/docs/?final
has a GFD number, and has "REC" in the document type
column. (The document process is described in GFD.1,
which you should read, if you haven't done so already.)

Plenty of OGF specs are proposed recommendations.
None have attained full recommendation status. I hope
that CDDLM will be one of the first.

To go the final mile towards getting the highest status,
you need to write an experimental (EXP) document (possibly
more than one, if it makes sense to treat each spec
separately) which essentially prove the case that there
are multiple, independent interoperable implementations
of the spec(s). This EXP document becomes input to a
formal review (which might involve external reviewers), 
on the basis of which the decision to award (or not
award) recommendation status is made. 

What "interoperable" means to some extent has to be localised
to the spec. I know from Steve Loughran that CDDLM has done
a lot of work on testing interoperability, so probably you've
done most of the work and just need to write it up. When you
do so, imagine yourself in the position of an external reviewer.
(Is the spec sound? Is it clear enough that a developer
can write an implementation that "interoperates" with others
from material in the spec alone? Do I believe that group has
done "due diligence" in this area? Do I believe that the
interoperability testing as described tests what should
and can reasonably be tested? Do I have any doubts about
the independence of the implementations?)

What some groups are doing is writing their interop plans
down in an INFO document before going about their interop
work. I think this is a good practice (and a good way of
getting more eyes on the approach to testing), but you're
probably already beyond this for CDDLM.

Most importantly, you should keep in contact with your
Area Director. That's Hiro.

Hope this helps,

Stephen

> -----Original Message-----
> From: Toft, Peter [mailto:peter.toft at hp.com] 
> Sent: 25 September 2006 17:13
> To: Hiro Kishimoto; Stephen M Pickles
> Subject: CDDLM-WG completion criteria?
> 
> Dear Hiro, Stephen:
> 
> As you probably know, I recently took over co-ordination of 
> the CDDLM-WG
> from my HP colleague Dejan Milojicic. Although I was involved in the
> inception of the group, I have been somewhat out of the loop for the
> last couple of years so I have some catching up to do.
> 
> What I want to check with you is what it means for CDDLM to 
> be 'done' --
> at least for the first pass at the standards. What we are currently
> working towards is an agreed specification set:
> 
> - CDL Language
> - Deployment API
> - Component Model
> 
> All of these are complete except for a resubmission of the 
> revised CDL.
> 
> To accompany each of these, we are running interoperability tests
> against three separate implementations of each of the specifications.
> The results of these tests will be captured in formal 
> documents, one for
> each specification, which will be submitted to the OGF editorial
> process.
> 
> So, given acceptable versions of each of these documents and the open
> source availability of some number of reference implementations, what
> else is needed to bring CDDLM-WG to a graceful conclusion?
> 
> Thanks and best regards,
> Peter
> 
>