RE: Issue #170: "Referencing Data missing from the message"

Hi Noah,

> Stuart Williams suggests:
> 
> >> The difference may not be so strong, given that Jacek [16] is open to
> >> the deserialisation process being lazy, avoiding the need to
> >> resolve/dereference references that are not actually required by the
> >> computation at the local SOAP node.
> >> 
> >> There may be a difference between Noah and Jacek on whether a SOAP
> >> Fault *MUST* be generated in the event that href
> >> resolution/dereferencing during deserialisation (lazy or otherwise) is
> >> actually made and fails.
> 
> 
> How can an external observer distinguish these two cases?

On the whole I'm inclined to agree... there is perhaps a question of which
black boxes an external observer looking at (some , but not all, failures to
resolve/dereference an href may result in externally visible activity)...
but that does feel like splitting hairs.

> Either can occur due to conditions known only within the processing node.

Agreed.

> In one
> case, I don't dereference because I was lazy and didn't try.  In the
> other case, I try, but there's a glitch in my network layer and I
> can't make the connection.  You worry that Jacek would REQUIRE that I
> fault only in the latter case, but how could you tell?  I might always
> claim:  "gee, actually, I never even tried to get at it."

'Worry' is probably not the right verb. I was trying to pinpoint the
difference in the positions that you and Jacek seem to hold. 

Jacek seems to be of the opinion that if an attempt to resolve/dereference
(and those I think are two different things) an href *actually* occurs and
fails a fault MUST be generated. 

You (and Henrik) seem to of the the opinion that in such circumstances a
fault MAY be generated.

I think that this is the only point of difference preventing closure of this
issue.

> 
> I think I'm still happiest with MAY fault.
>

I believe that that has been clear throughtout, and FWIW I too am happiest
with MAY, in part because there may be a significant lapse of time between
the arrival of a message and the dereferencing of any URI it might contain.
In such circumstances I think it would be quite difficult to prescribe quite
what to do with any generated faults... and on the topic of faults since we
generally only require that they be generated, the external observer test
also applies. How would an external observer know that a fault had indeed
been generated? :-)

> ------------------------------------------------------------------------
> Noah Mendelsohn                                    Voice: 
> 1-617-693-4036
> Lotus Development Corp.                            Fax: 1-617-693-8676
> One Rogers Street
> Cambridge, MA 02142
> ------------------------------------------------------------------------

Regards

Stuart

Received on Thursday, 3 January 2002 10:01:25 UTC