Re: Location, URI-header, etc.

Daniel DuBois writes:

	...

 > --- Implementation off, opinion on
 > On 200 OKs: Well, kind of.  Except the 'virtual places' it lists all point
 > to the same physical space that the document was provided from.  Speaking
 > from a 200 OK mindset, you already have the document, you don't really need
 > to go seek it out now (maybe later for new versions, but not right now).
 > The information in the URI: is for caches as I understand it.
 > 

So we need clear guidelines on what caches are supposed to do with
this information.  My current opinion is: if a http 1.1 cache caches
the document, it should be required to invalidate all other cached
items keyed by URIs from Location and URI-headers that came with the
document.  If the cache doesn't cache the document, it should be
recommended (but not required) that it invalidate items keyed by the
other URIs.  Also, since URI-header exists on 2xx responses, there's
no point throwing out Location on 2xx.  Their problems are identical.

 > >The URI-header may refer to multiple versions of a resource,
 > >but doesn't indicate which one is in the message.  
 > 
 > Since you say "in the message" I assume you refer to URI:s in 200 OKs.

Yes.

  The
 > URI item that doesn't vary is the one that explicitly is in the message,

Good idea.  Maybe this should be standardized.  (OK, keepers of the spec?)

 > according to our implementation.  The other URI item(s) are ways to get to
 > the information, but not as "directly".

That makes good sense to me.
	...
 > 
 > This is how I understand it.  But I could be SOOOO far off it isn't funny.
 > The whole URI/None acceptable area is a big fuzzball IMHO.
 > -----
 > Dan DuBois, Software Animal                          ddubois@spyglass.com
 > (708) 505-1010 x532                     http://www.spyglass.com/~ddubois/
 > 
 > 

It sounds like with your rule ("the URI that doesn't vary identifies a
way to retrieve another copy of the entity enclosed in this message"),
it all makes better sense.  So, shouldn't it be required that on 2xx
responses, one and only one of the URIs given must have no vary
dimension?

--Shel

Received on Friday, 1 September 1995 16:40:56 UTC