- From: Mark Nottingham <mnot@mnot.net>
- Date: Mon, 27 Aug 2001 07:33:54 -0700
- To: Yves Lafon <ylafon@w3.org>
- Cc: xml-dist-app@w3.org
On Mon, Aug 27, 2001 at 04:01:34PM +0200, Yves Lafon wrote: > On Sat, 25 Aug 2001, Mark Nottingham wrote: > > > > When you use POST to mean "MYFOOMETHOD" you sacrifice specificity > > > to get reuse of infrastructure. When you use POST to mean GET you > > > undermine existing infrastructure. And in my mind, caching is the > > > least interesting thing you lose. Hyperlinking and URI-based > > > addressing is way more important.... > > > > Agreed; I'd go as far as to say that no widely-deployed piece of Web > > software can or will ever be able to cache a SOAP message based on > > the current HTTP binding. I also agree that addressing is important - > > the fact that people are encouraged by SOAP to place multiple logical > > services on single URI is the primary reason why my stab at a SOAP > > caching module [1] is more complex than it needs to be. > > Note that, a POST reply can contain a Cache-Control header and then imply > a certain level of idempotence. The first POST will have a side-effect but > any subsequent POST with the same parameters won't. Of course you have the > possibility to cache the result as a bonus of this idempotence property. True. However, I don't know of any product that takes advantage of this, as it's considered impractical. (Curious - does Jigsaw?) > > However, I don't know that going to pure-GET is necessary to fix > > this; merely requiring a 1-to-1 service-to-URI mapping would do the > > trick. > > Usually jsp engines are single servlets, and the URI of the jsp is > passed as an argument to this servlet. You have then a 1-to-1 jsp-URI > mapping, same as what can be done for a SOAP engine to respect this. Yep. Or, just use query arguments. Cheers, -- Mark Nottingham http://www.mnot.net/
Received on Monday, 27 August 2001 10:34:13 UTC