Re: Idioms

[Danny Ayers <danny.ayers@xxxxxxxxx>]

On Sun, 10 Oct 2004 11:46:55 -0400, Joe Gregorio <joe.gregorio@xxxxxxxxx> wrote:

Personally I would like to see some more description of the problem in
abstract terms, before potential concrete solutions are evaluated.
Identifying patterns in the drafts seems like a good approach.

There is a wide spectrum for potential deployment scenarios - from
full WebDAV, through 8,7,6-verb REST to SOAPy POSTs. Format delivery
is at the edge with its heels dug in to GET + XML, but that only needs
one verb anyway. If anything, I suspect the extremes would actually be
the easiest to spec out, as the WebDAV and SOAP languages have their
own idiomatic styles which would constrain choices.

For 80/20 applicability, an optimal solution is likely to be that
which best fits the middle ground, services operating from the choice
of 2-6ish HTTP verbs, URI nouns and their descriptions in XML
documents. For compatibility and work-reduction reasons it would make
sense to try and draw from the Atom format as possible rather than
inventing syntax. But 'optimal' requires optimization. Even if it
isn't on the usual axis of performance, surely it would be better not
to do it prematurely.

Having said all that, I've no idea what would be the best way of
modelling this stuff. The kind of abstraction in Fielding's thesis etc
is pretty nice, I think it's Mark Baker who's expressed similar stuff
in RDF. From what I've seen of the WebDAV material much of the
background that needs to be covered has probably already been covered
there. But there are different constraints - whatever the protocol
looks like it should be possible to use it with the verbs moved below
the blank line.

Cheers,
Danny.

PS.  2 games in town - Annotea and InterWiki?

-- 

http://dannyayers.com