Re: Updated Working Draft: Enabling Discovery Of Alternative Representations

Paul Libbrecht writes:
 >  From reading such a document I get the impression that all negotiation 
 > logic should be concentrated before a redirect in order for caches to be 
 > enabled. Am I summarizing correctly ?

That's an interesting observation --- and I believe that it is
correct when one is talking about an on-device cache --- would
but it might not be the right thing for *every* cache for
instance, a mobile gateway might find that
"concentrating all negotation logic" as you put it is the right
thing since it might asume that all its accesses come from mobile
devices. On the other-hand, a caching firewall that serves both
mobile and desktop clients should certainly not preserve the
results of content negotiation since it would end up serving the
wrong representation in the following scenario:

0)    User-1 accesses http://example.com/info from a mobile
      device
1)    caching firewall negotiates with example.com
2)    Serves user-1 http://example.com/info/mobile
3)    Later, user-2 accesses http://example.com/info through the
      same firewall with a desktop device

Caching the results of the content negotiation performed for
      user-1 would be incorrect.
      
 > Which versions should a browser bookmark ?
 > 

That depends on what the bookmarked URL is used for ...

If the bookmark is *only* used as a bookmark and *only* in the
context in which it was created,it is probably safe for the
browser to remember the URL to which it was redirected. 

But in most cases, it is better to remember the "canonical URI",
i.e. the URL with minimal context encapsulated --- since that URI
can be emailed, passed to other application contexts etc. This is
also why I showed a 302 Redirect (content moved temporarily) in
my example.


 > paul
 > 
 > 
 > 
 > T.V Raman wrote:
 > > I've updated the document at 
 > > http://www.w3.org/2001/tag/doc/alternatives-discovery.html
 > > Comments welcome.
 > >
 > > ToDos:
 > >
 > > I'd still like to draw a nice generic schematic diagram in the
 > > final section showing how a canonical URI can have various pieces
 > > of context bound into it to create a multiplicity of URIs.
 > >
 > >   
 > 

 Thanks, 
 --Raman

-- 
-- T. V. Raman

Received on Thursday, 22 June 2006 16:49:49 UTC