W3C home > Mailing lists > Public > xmlp-comments@w3.org > January 2002

[Fwd: RE: Issue #12 closed]

From: Christopher Ferris <chris.ferris@sun.com>
Date: Tue, 15 Jan 2002 16:44:22 -0500
Message-ID: <3C44A2B6.4060909@sun.com>
To: xmlp-comments@w3.org
should have been copied to xmlp-comments not w3c-archive.

-------- Original Message --------
Subject: RE: Issue #12 closed
Date: Wed, 28 Nov 2001 12:02:55 -0800
From: "Larry Masinter" <LMM@acm.org>
Reply-To: <LMM@acm.org>
To: "'Christopher Ferris'" <chris.ferris@sun.com>
CC: <w3c-archive@w3.org>

I'm sorry for not replying sooner.

While the "Note:" is fine as it stands, I thought that the
Apache configuration for 500 error rewriting was merely
an example of the kind of problem one might encounter,
rather than the entire scope (although it's the only
one that I've documented). I remember hearing about other
kinds of "helpful" proxy servers that would also rewrite
error messages, but I could not find online documentation
for them when discussing this with the W3C XML protocol

So I would prefer a more generic warning:

   Note: There may be elements in the HTTP infrastructure
   configured to modify HTTP response entity bodies for
   4xx and 5xx status code responses; for example, some
   existing HTTP origin servers have such a feature as
   a configuration option. This behavior may interfere with

and that the 'recommendation' be turned into a warning,
since it may or may not be possible for the person deploying
SOAP services at a site to 'turn off' such behavior.

                            If the rewriting behavior
   cannot be turned off, SOAP/HTTP cannot be used in
   such configurations.

Perhaps instead of "Note:" it should be "Warning:".

 > -----Original Message-----
 > From: Christopher Ferris [mailto:chris.ferris@sun.com]
 > Sent: Wednesday, November 07, 2001 12:22 PM
 > To: Larry Masinter
 > Cc: w3c-archive@w3.org; Christopher Ferris
 > Subject: Re: Issue #12 closed
 > Larry,
 > Okay, here's what we came up with which hopefully will
 > address your concerns. We have crafted a NOTE (below) that we
 > will add to the specification in the appropriate place
 > in the HTTP Binding section where the HTTP response codes
 > are addressed.
 > 	Note: Some existing HTTP origin servers may be configured to
 > 	generate customized HTTP response entity bodies for 4xx and
 > 	5xx status code responses. This behavior may interfere with
 > 	the use of 4xx and 5xx status code responses carrying SOAP
 > 	fault messages in HTTP. It is recommended that such behavior
 > 	is disabled for resources accepting SOAP/HTTP requests.
 > Hopefully, the addition of this note in the spec will be
 > sufficient to assuage your concerns and allow us to close
 > issue #12.
 > Please let me know whether you agree.
 > Cheers,
 > Chris
 > Larry Masinter wrote:
 > > Chris, when I went to look, the Developmentor archives
 > > were there.
 > >
 > > The first issue was around the likelihood
 > > that HTTP implementations would rewrite 500 Server Error
 > > message bodies; I gave some specific examples of Apache
 > > configuration options that would do this when trying to
 > > implement SOAP with Apache and CGI, but pointed out claims
 > > that similar problems may exist with proxies.
 > >
 > > I know that some on the committee have discounted this
 > > as either unlikely or irrelevant ("don't use CGI and
 > > Apache to implement SOAP"), but unless this is the
 > > consensus, I would suggest at least noting the issue
 > > in the implementation notes for the HTTP binding.
 > >
 > >
Received on Tuesday, 15 January 2002 16:45:50 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:16:58 UTC