Re: generic MTOM proposal

On Nov 18, 2003, at 12:13 PM, [email protected] wrote:
>
> * A lot of the intro and context-setting from MTOM is missing, at least
> for now.  For example, it doesn't state clearly that typing 
> information is
> not preserved, and that the full DM is not transmitted.
>
> * There seems to be a mixture of DM and Infoset terminology.

Yes; I didn't know enough about DM to accurately specify things in the 
time I had. Also, out use of DM, as I understand it, is quite subtle; 
we're only using mechanisms that are unique to DM for optimisation 
candidates, whereas our handling of the rest of the document explicitly 
doesn't use typed information.

> * I think that to have a full proposal, we will need to do a 
> corresponding
> rewrite of MTOM, referring to Miffy where appropriate.  I think parts 
> of
> MTOM are still needed to describe the relationship to SOAP processing,
> possible new HTTP binding, etc.  I also think there would be an
> opportunity to lift some of the pertinent text from MTOM and adapt it 
> to
> Miffy, as I think that in some cases MTOM states more carefully what is
> going on with the optimizations (or if not, we should fix it.)

Agreed. I suspect that the resulting document would be quite small, 
because it's just an explanation of how to use MIFFY in the context of 
SOAP.


> I completely understand why Mark didn't tackle these things in a first
> pass, and I think he's given us an excellent start.  For the record, 
> I'm
> OK with the general Miffy direction, at least as something to explore, 
> if
> that's the will of the group, or I can easily live with a SOAP-only
> solution for now.  I am a bit concerned about how we balance
> investigations of Miffy with David's suggestion of MTOM last call 
> coming
> out of the F2F.  I think that getting some of the MTOM details to last
> call state will be a bit tight even without Miffy, though we can 
> probably
> do it with some effort (also on my to do list is to try to respond to
> David's call for outstanding issues.)  I don't see how we can do a 
> proper
> Miffy refactoring to Last Call quality by the F2F.  Even if we could,
> there would be something to be said for gathering comments informally 
> for
> a week or two, to at least get a sense that we have met the concerns of
> those who wanted a more general optimization scheme usable outside of
> SOAP.

Agreed, although I question whether Last Call by the F2F is a 
reasonable or necessary goal. Generally, I'm all for getting this done 
quickly, but a few extra weeks won't make much of a difference, surely, 
especially if the content is much more modular and easier to 
understand?


--
Mark Nottingham   Principal Technologist
Office of the CTO   BEA Systems