- From: Oisín Hurley <ohurley@iona.com>
- Date: Tue, 14 Nov 2000 17:24:07 -0000
- To: "Williams, Stuart" <skw@hplb.hpl.hp.com>, <xml-dist-app@w3.org>
[deletia] > For straighforward unidirectional XP message delivery this requirement is > fine. However, for matching 'synchonous' request/response pairs I > think this > requirement implies either: > > a) the existence of a transport independent mechanism > 'addressing' responses > and for matching request response pairs (ie. one that works say > over SMTP as > a transport as well as HTTP). Yes. > b) a transport specific mechanism for 'addressing' responses and matching > request/response pairs that gets pushed downward into the definition of a > transport binding for a given 'transport protocol'. No. > I think that if neither of these holds then we loose the neutrality this > requirement is intended to "ensure". > > Have I missed something? The implication for the coverage of RPC-style messaging is that there is information within the message model (possibly in terms of a mandatory header) which contains some device to perform request/response correlation. If the transport protocol in use happens to have such an artifact as part of its makeup, then the XP mapping to that transport protocol is free to use it as the expression of its correlation device. It is also quite free to refuse to make use of the built in correlator and explicitly deposit correlation data on the wire. The important thing is, whatever it does for that particular protocol should be consistent! cheers --oh -- ohurley at iona dot com +353 1 637 2639
Received on Tuesday, 14 November 2000 12:28:12 UTC