Implementation report for content-language
	Submitted by Harald Alvestrand

This report considers implementations of the content-language and
accept-language headers described in
draft-alvestrand-content-language-02.txt, which is an update of this
part of RFC 1766 (which was a Proposed Standard).

A lot of the information in this note is drawn from the HTTP/1.1
implementation report, currently online at
http://www.ietf.org/IESG/Implementations/http1.1-implementations.txt

Feature list
------------
The following features of draft-alvestrand-content-language have been 
identified:

- Content-language header
- Accept-language header with Q values
- Accept-language header without Q values

The implementation report for HTTP/1.1 lists the following implementation 
statistics:

Clients   |Servers   | Proxies  |Feature
 7t 2y  4n|10t 3y  5n| 4t 1y  3n|H 14.4 Accept-Language
 5t 3y  5n| 7t 6y  5n| 3t 2y  3n|H 14.12 Content-Language

Where "t" is tested, "y" is implemented but not tested with another 
independent implementation, and "n" is not implemented.

The HTTP/1.1 implementation report does not specify the use of Q values, 
since that feature was undocumented in the HTTP/1.1 specification.

Internet Explorer 5.5 implements Accept-Language with Q values.
Opera 5.12 implements Accept-Language without Q values.
Netscape Communicator 4.76 implements Accept-Language without Q values.
Lynx 2.8.4dev.11 implements Accept-Language, but does not syntax check the 
user's input, so any style of Accept-Language is possible.

The Apache web server (1.3.20) supports both styles of Accept-Language.

The PMDF and iMS MTAs support Content-Language headers in email; they both 
generate them, accept them and are able to take action on them.
Parsing of Content-language is a required feature of RFC 2060 (IMAP4rev1).

I believe this collection is enough to support advancing this specification 
to Draft Standard.