ietf
[Top] [All Lists]

Re: Comparison of ICAP and SOAP

2001-06-27 05:50:03
I'm not sure why we're doing this on the ietf discussion list, but it's
clear they have nothing better to do....even so, I suggest we move this to
the OPES list.

So, SOAP could be used to implement ICAP, but then again so could
BEEP.

There is a SOAP over BEEP proposal.  SOAP provides the RPC binding semantic
and BEEP provides the transport.  We may need to make sure our requirements
are properly addressed, but that's business as usual.

Not too many of the ICAP people are interested in using SOAP,
though, as their requirement is to allow 'wire-speed' vectoring and
processing, and they find the overhead of XML unacceptable.


I haven't taken a poll so I can't comment on how many of the icap
implementers are rethinking their implementations.  When we polled people at
the last OPES workshop, there were only a few people who knew anything about
SOAP.  Thus my previous assertion that icap was created in isolation (and is
a flash in the pan).  I assume that as rational IETF participants, when we
do a careful comparison and as people learn more about it, we'll see some
movement of opinion.

But, I can state that there is very significant support for SOAP-based
solutions in the industry.  SOAP applies when a web service (e.g., an web
intermediary) needs to call on another for services (e.g., content
adaptation).  For example, a service that provides language translation
could be called from an origin server or from an intermediary or, for that
matter, directly from a client.  We should have two (or more) bindings for
that service?  Seems silly to me.  And if you really want edge services to
be a thriving part of the enterprise or ISP market places, seems like you
should use the service bindings that other web services are using.

As for XML overhead, I'm not sure what overhead there is?  One can build
syntactic routines that optimize for the supported functions if there are
concerns about the soap envelope structure.  However, from what I've heard
and from my observations of the icap formats and protocol the real concern
was code reuse.  They wanted to be able to use existing (or nearly
unmodified) web servers.  Well, that may have been fine for a tactical
solution, but the industry is using the web services infrastructure
(SOAP/XMLP, UDDI, WSDL).  OPES should do so as well.

BTW, since the icap forum has found another standards body to rubber stamp
their work, I think we can move on (except for whatever useful development
experience might have been gained in the process).  They get their marketing
literature and we get to do something useful for the industry.

P.S.  I normally work in pairs, I play the bad cop...someone want to pick up
the good cop role?  Cheers.

Lee M. Rafalow
Voice: (919) 254-4455, Fax: (919) 254-6243
IBM Internet Technology Management
IBM Corporation
P.O. Box 12195, BRQA/502
RTP, NC 27709 USA
Alternate email: rafalow(_at_)us(_dot_)ibm(_dot_)com
Personal email: lrafalow(_at_)mindspring(_dot_)com



<Prev in Thread] Current Thread [Next in Thread>