- From: Mark Baker <distobj@acm.org>
- Date: Tue, 30 Sep 2003 12:02:44 -0400
- To: David Orchard <dorchard@bea.com>
- Cc: www-ws-arch@w3.org
Yes, SMTP is stateful, as are many other application protocols. If the reason this was brought up was to suggest that statelessness should not be an SOA constraint though, keep in mind that SMTP sessions (not connections) include only a single mutation (i.e. one DATA invocation), thereby avoiding the partial failure problems typical with stateful connectors. Also, between sessions, SMTP is stateless. So it's very atypical, and therefore a poor example. An architectural style is only a guide, not a conformance or interoperability statement. What matters most is that architects are familiar with the properties of the style, and that they understand what properties are sacrificed when constraints are not followed. Stateless services have several superior properties to stateful services (security, visibility, reliability), without much of a downside (larger messages) for this space. For this reason alone, I feel that statelessness (technically, stateless connectors) should remain a constraint. Also, I've heard that you personally Dave, see self-description as pretty key to Web services. I'm sure I don't have to tell you that if you're using a stateful connector, you are by definition, not self-descriptive. Mark "will not mention REST" Baker. On Fri, Sep 26, 2003 at 09:03:01PM -0400, David Orchard wrote: > > I was reading through rfc822 and rfc821 today. SMTP seems to be a stateful > specific method protocol that has been widely and sucessfully deployed on > the internet. Heck, there's even state transition diagrams. Anybody > disagree? Please make sure to have read smtp before responding. > > Cheers, > Dave -- Mark Baker. Ottawa, Ontario, CANADA. http://www.markbaker.ca
Received on Tuesday, 30 September 2003 11:58:44 UTC