Lists Home |
Date Index |
Mike Champion wrote:
> On Wed, 12 Mar 2003 14:09:18 +0000, Bill de hÓra
> <firstname.lastname@example.org> wrote:
>> The thing is, to do anything interesting, are the users being burdened
>> with reinventing protocols? As Geoff Arnold pointed out recently, API
>> and protocol design require very different mindsets and approaches.
>> Just shipping data around, really means just shipping data around
>> /with/ service level agreements.
> To be honest, this is something I'm having a hard time coming to grips
Ok, but you just need to see the way the industry is hovering around
a nunber of reliable messaging standards, recent discssion on the
W3C ws lists, or at a lower level in Java-land, the SOAP --> Servlet
--> JMS idiom, or even SOAP-->ASP-->BizTalk, to realize that
shiipping stuff around is a minimal expectation. Assuming the apps
can talk, the next thing people want is the data Fedexed.
> The RESTifarians make a strong point that every WSDL file defines
> a new "protocol", and that ordinary users can't be expected to grasp the
> subtleties of protocol design. Getting back to my original point, that
> seems like saying that the specific message format that a CGI (or
> ASP/JSP/etc.) backend expects is a "protocol."
I don't buy the ordinary user excuse too much. But I do know it
takes time and effort to craft anything halfway decent in terms of a
protocol or MEP above beyond request-response, which contradicts or
offsets to a degree the application integration speeds WS are
supposed to offer. Maybe the tradeoff is still a good one. But now
we're way of the XML chart...
> True in the literal
> sense of the word "protocol" in English, but I'm not sure it means
> anything more than "the minimal expectation by the server of the
> information content needed to do its job" (which is my understanding of
> the position that folks including Walter Perry and Sean McGrath
Which is entirely sensible. But we're talking about sequences of
messages that don't compose so well from request-response pairs.
Reliable sequences can require 5 or more messages for the endpoints
to come to agreement. Nobody talks about tranasction-neutral XML
databases; maybe that's more obviously dubious.
Bill de hÓra