[Nmc-wg] Review of documents

[Freek Dijkstra]

Hi all,

My apologies for the (very) late reply.

I have a few generic comments and questions on "Extensible Protocol for
NMC" for now and will follow up with some more detailed remarks later.
(I planned to have the detailed remarks now, but was not able to finish
reading all I wanted).

My main point of criticism is that I do not (yet) understand why this
protocol was developed; as a naive reader it seems to highly overlap
with what has been done with WSDL. Unfortunately, I'm not an expert on
either, so I might miss the point. WSDL was developed as a generic
request-response protocol framework, and so is "Extensible Protocol for
NMC".

I still like the work, for example the idea on metadata is good. I just
recommend to (a) build more on existing frameworks (e.g. specify that
the protocol MUST use webservices instead of SHOULD, and refer to
WS-whatever for authentication extension) and (b) make it more clear
what was added.

While I'm not very familiar with RELAX-NG, I think it is a good way
forward, as it can be translated to XSD, and hopefully, this means there
can be automated syntax check for messages. Has this been done so far?

Last, the wording could be better in some places, especially when it
comes to the RFC 2119 words (MUST, SHOULD, MAY). Two examples: "MAY or
MAY NOT" (page 6), and my favourite: "Data MUST contain information,
especially if it was empty in the initial request". Beside the
obviousness of this statement (what else then information?) and the
unclarity where "it" refers to, it seems that "SHOULD" was intended
here, not MUST (why else the "especially"?).

Summarizing:

- The ideas are good
- It should build more on WS-stuff (especially since this is the
  OGF, which is quite WS-frantic if you aks me).
- The wording could be more strict.

I hope to have more detailed replies while I finished the reference
document, and read a bit on WSDL. (the awful truth is that I have not
programmed a webservice yet).

Regards,
Freek