RE: i24: Requiring Allow in 405 responses

Julian Reschke wrote:
> Therefore, if we allow the server not to include all methods, we also 
> have to allow a client not to believe what it got, thus:
> 
> "However, the indications given by the Allow header field value SHOULD

> be followed."
> 
> should be dropped then.

That needs to be dropped anyway. It cannot be implemented unless the
client caches the Allow headers it receives, but there is no
(SHOULD-level) requirement for any client-side caching in HTTP. Besides,
there is no freshness information associated with the Allow header to
control the caching. (Maybe that is why it was defined as an entity
header?)

Plus, besides wasting the client's time and a little bandwidth, there is
really no harm in ignoring the requirement, and no benefit to following
it. It might be different if extensions to HTTP are being used, but
those extensions can seperately document their extended usage of the
Allow header as they see fit. 

- Brian

Received on Monday, 3 March 2008 13:23:35 UTC