REST and POST

Mark Baker distobj@acm.org
Thu, 23 Aug 2001 13:48:59 -0400 (EDT)


> I have a sincere objection, but not a fundamental one.
> 
> The syntactic properties of GET and POST are different, as well as their
> semantic ones.  It must be possible to use POST semantics for interactions with
> no side effects -- otherwise implementors who need POST syntax will continue to
> ignore the semantics, and the semantics will continue to be meaningless.

How about an alternate way of doing what they need with GET (e.g.[1])?  There
are good reasons for ensuring that GET remains the only method with which a
representation can be returned without side-effect a priori (i.e. if sometimes
the POST left a side-effect, and sometimes it didn't, then POST is the right
method).

TimBL's web axioms talks about this (though I'm still undecided whether
this one should count as an axiom - it seems more a practical issue to me,
not a fundamental one - could be wrong though);

http://www.w3.org/DesignIssues/Axioms.html#state

Which reminds me that we should integrate this content into the Wiki
somehow.  I'll add it to the RestWikiTodo.

> This was the stumbling block for most of the early conversation about REST/RPC.
> It can be dealt with and put away or stay on indefinately as a nagging problem.
> 
> Unless someone can show why POST/GET syntax have to be connected on a formal
> level with side effects...

See [1] as an example of doing complex queries over POST.

 [1] http://lists.w3.org/Archives/Public/www-rdf-interest/2001Aug/0182.html

MB