Re: Last-Modified in chunked footer

On Fri, 5 Sep 1997, John Franks wrote:

> There is a set of headers which clients need to have before the
> entity body arrives.  There is a set of headers which servers 
> can only calculate after the entity body has been served.  
> 
> The good news is that all the evidence so far indicates that the
> intersection of these two sets is empty.
> 
> The bad news is there does not seem to be a one sentence
> characterization of either set.
> 
> Most likely we will have to enumerate one set and say that the
> other set is its complement.  I see no clear case for picking one
> over the other.

Where it makes a difference is for headers that cannot be enumerated,
i.e. headers not defined in the HTTP spec.

If a client can ignore all (or all except some?) header fields sent in
footers and still be compliant, then I think it doesn't really matter.
But there should then be some note warning server implementers about it.

Btw, the current BNF allows only entity-header fields in footers.  "Vary"
is not classified as an entity-header, so this would also have to change
if Vary should be allowed.

   Klaus