Re: Accept-Transfer header field (was HTTP/1.1 Issues: TRAILER_FIELDS)

At 16:48 11/18/97 PST, Jeffrey Mogul wrote:
>Henrik Frystyk Nielsen writes:
>
>       Accept-Transfer  = "Accept-Transfer" ":"
>                           1#( t-codings [ ";" "q" "=" qvalue ] )
>
>Shouldn't that be:
>
>       Accept-Transfer  = "Accept-Transfer" ":"
>                           #( t-codings [ ";" "q" "=" qvalue ] )
>
>since one of your examples is:
>
>       Accept-Transfer:
>
>(I know, I made the same mistake for Accept-Encoding in rev-00.)

duh - yes, you're right

>Also, since (as Roy has pointed out) the requirement for protecting
>Accept-Transfer with Connection makes requests somewhat verbose,
>perhaps we should be using a shorter name ... "Accept-Trans"
>would save 6 bytes per request.  
>
>I'm not even sure this header should be called "Accept-anything",
>since it's a hop-by-hop mechanism and thus pretty much orthogonal
>to content negotiation.  Maybe "OK-Trans" (saving another 8 bytes
>per request)?  It's not as if any human being is supposed to be
>reading these headers.

The only reason for using an accept-* name is that its BNF falls into that
category of header fields. The same could be said about Accept-Encoding
along with a long list of other problems, but that is too late to fix now
(see [1], for example).

What about calling it Accept-TE - that saves 12 bytes.

>P.S.: OK, I *do* include HTTP implementors in the set of human
>beings :-)

I think that's a different discussion altogether ;-)

Henrik

[1] http://www.w3.org/Protocols/HTTP/Issues/#ENCODING_NOT_CONNEG
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk

Received on Wednesday, 19 November 1997 09:53:32 UTC