RE: Issue #12: HTTP Status Codes 500 v 200

Based on the discussion that took place on this thread [1] prior ot the June
F2F I prepared the attached presentation for the F2F meeting. Unfortunately,
this issue didn't get discussed at the F2F, so I'm posting this material
here for information.

The presentation includes what I think are probably the only three possible
resolutions to this issue (no doubt someone will prove me wrong :->):

1) Use 2xx status codes for ALL XMLP/SOAP messages carried in HTTP response
messages

2) Use 5xx status code for ALL XMLP/SOAP Fault messages in HTTP response
messages (and 2xx for ALL non-Fault XMLP/SOAP messages) (status quo I
think).

3) Use 2xx status code for some (TBD) classes of XMLP/SOAP Fault message and
5xx for some (TBD) classes of XMLP/SOAP Fault message carried in HTTP
responses.

During the course of discussion, I was also referred to an Internet Draft
(now expired) by Keith Moore on the use of HTTP as a substrate for other
protocols [2]. Section 8 of that draft contains suggested guidelines
appropriate to this issue (extracted in the attached presentation). 

My reading of Keith's 2nd and 3rd guidelines do not appear to rule out
either 1,2 or 3 above!

Keith's 4th guideline is a word of caution about the possibility of HTTP
intermediaries altering the content of HTTP error responses.

Best regards

Stuart Williams
HP Labs, Bristol, UK.

[1] http://lists.w3.org/Archives/Public/xml-dist-app/2001Jun/thread.html#15
[2] http://www.normos.org/ietf/draft/draft-moore-using-http-01.txt