Re: Web-friendly SOAP

Walden Mathews wrote:
> 
> If SOAP and "web services" represent an evolutionary step from using
> the web interactively (browser and human user) to the less constrained
> model including non-human agents, then doesn't this strongly imply
> a corresponding UDP binding for HTTP? 

Could you make the logic you are using there a little bit more explicit?

Is it something like:

If asynchronous, best-effort delivery is necessary for web services then
it is probably also use for "the web" and should therefore be provided
at a level that would make it available to both of them? That makes a
certain amount of sense.

> ... Is there anything unRESTful
> about asynchronous or connectionless state transfer? 

I don't believe there is anything unrestful about HTTP over UDP. *BUT*
let's say I send you an XML document and it has three hyperlinks in it.
One is to a stylesheet. One is to some form of semantic description. One
is to an inclusion (perhaps XInclude). In order for you to process the
document you may need this inclusions. So at that point you need a
request/response MEP. So sending things over UDP is just fine
(especially for performance), but any node that can't download data from
the Web using HTTP isn't really on the Web and should not expect to
share in the benefits OF the Web.

 Is HTTP/UDP perhaps
> a key missing piece?  Is there discussion on this that anyone out
> there could point me to?

 * http://research.sun.com/techrep/1999/abstract-71.html

Note that there are two different ideas that could be called "HTTP over
UDP". One is unidirectional REST operations over UDP and the other is
truly HTTP over UDP (i.e. including the request-response model).

 Paul Prescod

Received on Wednesday, 19 June 2002 13:37:12 UTC