Re: Article on WS architecture and best practice ... may be of in terest

Cutler, Roger (RogerCutler) wrote:
> Paul -- I would find it very useful if you amplified what you say on the
> bottom of this thread a bit.  What factors do you have in mind that would
> make a web service interface a "good network application".  Does this have
> to do with the granularity of the calls because of latency considerations?
> Or the nature of the argument typing -- that is, preferring interoperable
> typing to, say, Java-specific types?  Am I on the right track or are you
> thinking of other things entirely?

Those are two big ones. You need to be able to deal with network 
failure. You want to rely much more heavily on standards and generic 
interfaces. For instance I've been asked recently about how weblogs talk 
to each other. One proposal out there uses an XML-RPC method call. My 
proposal is to POST an RSS. Use a standard XML syntax over a standard 
method on a standard protocol. I think that good networked applications 
are as stateless as possible and tries to decouple the processing model 
of the various participants from each other.

I'd like to say that networked apps should make more stuff addressable 
but in most local programming languages most objects are addressable 
most of the time so I guess you can't do much better than that.

There are probably other things I've forgotten. Good references that 
should back these up include:

  * http://java.sun.com/people/jag/Fallacies.html
  * http://research.sun.com/techrep/1994/abstract-29.html
  * http://www.ics.uci.edu/~fielding/pubs/dissertation/top.htm
  * Adam Bosworth's various articles (I don't always agree with them but 
he's got a lot I do agree with)

  Paul Prescod

Received on Monday, 7 October 2002 19:19:05 UTC