[
Lists Home |
Date Index |
Thread Index
]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
> -----Original Message-----
> From: Mike Champion [mailto:mc@xegesis.org]
> Sent: 12 April 2002 22:45
> To: 'xml-dev'; Bill de hÓra
> Subject: RE: [xml-dev] Will Web Services Kill HTTP?
>
> Mike:
> I'm basically with the REST people on this issue -- the Web
> is the most successful
> distributed application ever built, and SOAP can't have it
> both ways: either
> leverage the architecture that makes the web work, or don't
> expect the "SOAP Web"
> to be as robust and scalable as the HTTP web is without a lot
> of thought,
> experimentation, investment, and time. SOAP has the
> mindshare at the moment, but
> the Web has actual reality behind it. I love this quote from
> THE MYTHICAL MAN-
> MONTH: "The real tiger is never a match for the paper one,
> unless actual use is
> wanted. Then the virtues of reality have a satisfaction all
> their own."
In fairness the technical people driving the SOAP standard see the
distinctions. The main argument against REST for application comms,
as opposed to application-people comms in my mind is to do with
data typing. People are notoriously robust in the face of untyped
data, software less so. I'm still learning, so maybe it's a
non-issue, but REST comes across as untyped. I haven't heard any
killer arguments against REST based on data typing, but it seems
important.
<aside>
I was trying to figure out what a REST setup would look like in a
non-distributed system, particularly Roy Fielding's notion of a
single interface to all resources: if REST is so cool maybe we can
steal some ideas and build better software period. An interface
with three or four verb methods that used uids as an argument
instead of computerDeutschMethodCalls() crossed my mind, something
that's come up on this list not so long ago. Maybe the base
language could be sanely extended using lambda or macros (danger: I
don't much about lambda or macros :). The idea of using few but
well known verbs is speech act theory for the rest of us, but you
see thr idiom in everyday OO code as well as get/set methods. It
would be nice to try and connect HTTP verbs with the speech act
theory being done in the agents community, or define an object
system in terms of 4 or 5 methods, just to see what would wash out.
</aside>
> What I really find strange is the conjunction of this article
> and the one that
> Simon pointed to. It does not take a totally pathological
> degree of paranoia to
> wonder if all this is being driven by folks who want to sell
> us all yet another
> shiny new networking infrastructure to replace the HTTP-based
> one that we all
> bought over the last 6-8 years ... That's nice, I'm all for
> stimulating the economy
> and getting our collective stock options back above water,
> but we must remember
> that the HTTP web was not a paper tiger created out of
> standards committees and
> marketing hype and visionary whitepapers, but a real tiger
> that came out of the
> wild and devoured those who stood in its way.
Well you know...if I had all the corporate mail server space locked
up along with the identity servers, I might be interested in
getting off HTTP too, it might help me get around annoyances like
Apache and Linux :) Anyway if you buy Geoffrey Moore's theories on
technology markets, infrastructure swapouts are how we make money,
and kind of inevitable; tho' swapping the web out is pretty bold
stuff.
Bill de hÓra
-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.4
iQA/AwUBPLhA0eaWiFwg2CH4EQIwqgCeLiufHg88u8qHAfRnKZ0UY45fjYkAmgIj
+vwDm7s6OQwqDIceGqthx9ub
=L8Op
-----END PGP SIGNATURE-----
|
