[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Date | Thread | Author

[XML-SIG] SOAP Skepticism


> Well said, and I made this poing to Uche several months ago. But I've
>> since realized that this has nothing to do with web services...i.e.
>> integrating organizational processes through the web. If it way too
>> expensive to use a technology optimized for point-to-point, like RPC,
>> rather than a technology optimized for creating a shared and common
>> namespace and programmatic interface like HTTP.    (01)

Paul,    (02)

Following the links you posted to the list was an interesting read indeed
and started a couple nice discussions here at Logilab (ask Alexandre ;-).
Instead of all your arguments about caching, I think it would be better to
emphasize the fact that, as Uche or you said, you cannot rely on the
closed world assumption when dealing with the internet. Providers and
users of web services and web resources are losely coupled and that makes
the message-passing paradigm perform far better than the procedure/method
invocation one. This means that you always know how to pass a message,
even if you may not be able to figure out how to deal with it, which is in
my view better than "we don't even talk the same language !".    (03)

And HTTP/FTP/SMTP/NNTP requests/responses that enable the growth of the
internet are very much like messages aren't they ?    (04)

My point is that loosely coupled systems work best with messages, whereas
tightly coupled system can benefit from remote invocation mechanisms
because they pass pointers/references around.    (05)

What's good for the inside is not necessarily good for the outside. It's
like programming scientific applications that use parallelism, you call
functions on the same node and send messages to other nodes.    (06)

HTTP is a protocol to efficiently pass "messages" around (BTW, wouldn't a
MOM relying on HTTP be a nice thing to have ?), but RPC/RMI/SOAP is not,
hence HTTP is a better fit for interconnecting systems over the internet.    (07)

As for the "RPC/RMI/SOAP provides type/call checking much earlier than
messages over HTTP do" argument I read in the KnowNow mail archive, I
don't think it can be of much appeal to a Python fan like myself :-)    (08)

Anyway, I didn't want to resume the xml-sig flooding and I'll be happy to
take this off-list with other interested parties.    (09)

-- 
Nicolas Chauvat    (010)

http://www.logilab.com - "Mais oł est donc Ornicar ?" - LOGILAB, Paris (France)    (011)