Re: i70: cacheability of status 303

Mark Nottingham wrote:
> Any other thoughts on this? It's a fairly substantial change. David B 
> suggested a few re-wordings, but I don't see agreement around those yet.
> Personally -- there are a few editorial changes that I'd make, as the 
> proposal is a bit wordy;
>    - "in order to obtain a representation corresponding to the response, 
> be redirected again, or end with an error status" is superfluous, as is 
> "generally" and "by Cache-Control or Expires header fields."
>    - "...transferred by the server over HTTP" seems clunky, as does 
> "follow-on", although I don't have any immediate suggestion for either.

+0 -- that is, it is wordy, but I don't think this is a problem.

> Also, we seem to be implicitly dropping the note to implementers WRT 
> HTTP/1.0 support for 303. Are we confident that this is no longer a 
> problem?

I would lean to "no, it's no problem", unless somebody can point to a 
client that still is in significant use and has that problem. Let's try 
to get rid of *some* of the historical stuff :-).

> Finally, 2616 makes it a SHOULD-level requirement that the Location URI 
> is a different URI. Are we dropping that explicitly? One easy fix would 
> be to change "The Location URI indicates a resource..." to "The Location 
> URI SHOULD indicate a different resource..."

The requirement is still there ("The server directs the user agent to a 
different resource, indicated by a URI in the Location header 
field..."), it just doesn't use RFC2119 terminology to say that. I 
personally have no problem with that.

> Otherwise, it looks good to me. I've copied the original spec text below 
> for comparison.

Best regards, Julian

Received on Tuesday, 13 November 2007 14:53:20 UTC