- From: Danny Ayers <danny.ayers@gmail.com>
- Date: Wed, 8 Jun 2005 23:57:05 +0200
- To: Mark Baker <distobj@acm.org>
- Cc: public-web-http-desc@w3.org
On 6/8/05, Mark Baker <distobj@acm.org> wrote: > > Looks good, Danny. I was thinking of use cases at a higher level (i.e. > avoiding the message trace, at least for now 8-), but that still > captures what I think is important I went for that level a little purposefully, when abstracted out the obvious can get hidden. Compare the approach to rich RSS posts with XML-RPC in the MetaWeblog API with straight RSS over HTTP.I don't imagine anything that gruesome is likely to appear in these parts. But the idea of using XPath to describe parts of messages has been raised.Pondering this it occurred to me that the response to "2+2 ?" is better either as "2" or something expressed in a format about which the parties have knowledge (e.g. <result>2</result>). Things like: result:value="/result/text()" result:type="xsd:integer" add complexity without any real benefit. (though you might want to check your > math on that first one 8-). Blasted slide rule... > FWIW, those are the kinds of use cases that forms languages are > targetted at addressing. Ok, thanks. I must admit I still haven't completely grokked the forms approach. > Does anybody else have some use cases supporting code generation > requirements? Client and server skeletons for the adding might help me get a working calculator... One point I want to slip in somewhere, the support of unlimited service-service hops is maybe an assumption that maybe should be made explicit, given that many current services have terminal nodes at either end. Cheers, Danny. -- http://dannyayers.com
Received on Wednesday, 8 June 2005 21:57:09 UTC