Re: Vary and Accept-Language

> In HTTP 1.1:
> If the server sends Vary: Content-Language must it also
> send a Content-Language header ?  (I would think so)

It's clear from the example later, you meant "Vary:
Accept-Language". And no, there is no requirement that
Content-Language be provided even though the response
includes "Vary: accept-language".

> If the server has more than one language available, but the user agent
> does not send Accept-Language, need the server send Vary:
> Content-Language at all ?

The server must send "Vary: accept-language" if and only if the
response would be different if a different accept-language would be
sent.

> If it does, what is the action of a proxy server on getting a request
> for that URL without an Accept-Language header ?

"Accept-language not supplied" is a different request from
"Accept-Language: en" if the response says "Vary: accept-language".

> How about:
> URI=index.var, server preferred language is English
> User Agent sends        Server sends           Proxy Server caches
>  -                       english.html          URI=index.var
>                          Content-Language: en  Content=<english.html>
> Accept-Language: en      english.html          URI=index.var
>                         Content-Language: en  Content=<english.html>
>                         Vary: Accept-Language Content-Language=en
> Accept-Language: fr      french.html           URI=index.var
>                          Content-Language: fr  Content=<french.html>
>                          Vary: Accept-Language Content-Language=fr

No, this is wrong. It's much simpler than this. If there's more than
one language, and the response would be different if you send a
different Accept-language, then the response must say "Vary:
Accept-Language". It's also a good idea to send Content-Language to
identify the language you actually sent, but there's no requirement
that you send it, that the Content-Language actually match (any of
the) Accept-Languages sent, etc.

It is my belief that my responses are the only reasonable
interpretation of what draft-ietf-http-v11-spec-06 actually says.  Do
you disagree?

Larry

Received on Monday, 22 July 1996 15:27:19 UTC