RE: Proposed issue: site metadata hook (slight variation)

 
> But there's a crucial difference. My proposal (which is really only a 
> minor variation on Patricks) uses a _request_ header, to 
> allow clients 
> to actively request metadata. The current 
> distributed-registry proposal 
> only allows servers to provide unsolicited metadata responses.

The problem is that if you are using a GET with a Meta: header, 
then are you really getting a representation of the resource?  It 
seems to be stretching the semantics of the GET method. 

Also, how do you PUT/DELETE/POST to that metadata, just using 
the Metadata: header? It seems morally equivalent to a set 
of M(XXXX) methods which has the same problem.

The only issue with having the OPTIONS method return 
a Meta-Location: header is that it takes multiple requests 
to obtain the data, an OPTIONS request to get the uri of the 
metadata, and a GET on that uri to actually obtain it.  I see
this as both beneficial and necessary.

Really the only difference between this mechanism and the one 
proposed by Costello et al [1], is that it uses the OPTIONS method to 
bind the resource to it's metadata and thus avoids the issue that 
Fielding brought up of always having the extra header. [2]


[1] http://www.xfront.com/dist-reg/distributed-registry.html
[2] http://groups.yahoo.com/group/rest-discuss/message/3315

Received on Wednesday, 12 February 2003 10:15:29 UTC