XOM API for Processing XML with Java

[Steve Loughran <steve.loughran AT gmail.com>]

On 7/7/05, Elliotte Harold <elharo AT metalab.unc.edu> wrote:
> John.Cowan wrote:
> 
> > I think the sensible approach would be to specify a profile of XHTML to do
> > this.  Browsers would cope automatically, and RESTful web services would 
> > be
> > able to do something sensible with it.
> 
> 
> RDDL might be a good example of how to approach this.

mmm. good point. maybe I should create a new spec, in the group of
REST-* protocols, then create a working body REST-I to verify
conformance :)

> 
> Also, if the server is hosed enough to return 500 series errors you
> might not be able to count on it being able to serve a stylesheet
> referenced by an xml-stylesheet processing instruction.

another good point. Over in SOAP-land, a 500 series error is what you
are meant to generate to indicate a fault (it is mandatory, though
WS-I say receiving a 200+fault should be interpreted as a fault too).
The problem is
 -500 has lots of meanings. You have to look for the soapfault
 -some stacks are dumb.

Back when I was doing XMLRPC work, we got a support call from our
remote customers complaining "we were sending bad XML". We eventually
got hold of the XML in question, and  found out that it wasn't bad
XML, it was valid HTML, an IIS 'there is no page of that name here'
error. And we were running apache.

Somehow their configuration was pointing to the wrong host, so instead
of getting XMLRPC responses of 200+text/xml+XML, they were getting
404+text/html+HTML content, and their dumb little stack wasn't looking
at the obvious cues before handing it off to parsing.

I think the moral there is never underestimate the incompetence of the
far end, and if you are the far end, dont underestimate the
incompetence of the other end either. And never try parsing HTML in an
XML parser, unless it is XHTML.

-S.