Language negotiation and robots

Language negotiation seems to work quite well (at least for
Western languages) between eg. the Apache server and
Mosaic-l10n, Netscape 3 and (I presume) Lynx 2.5).

A description of this may be found at 
http://vancouver-webpages.com/multilingual/howto.html
and elsewhere.

I am interested in the interaction between this and web robots; 
specifically, how to tell the robot that the document is available
in more than one language.

Using the Apache type-map model, if no Accept-language
field is sent, one gets the first URL (assuming qs is not used or are equal).
Would it be valid to do something like this:

URI: start; vary="language"

URI: unknown.html
Content-type: text/html
Content-language: en,fr,de

URI: german.html
Content-type: text/html
Content-language: de

URI: english.html
Content-type: text/html
Content-language: en
etc.

so that a robot with language-accept unset would retrieve the 
"Content-language: en,fr,de" header, and then retry with 
Accept-language set to en, fr, de in turn?

As this multiple-language header seems to be valid in the current draft 
spec., I presume it should not break (future) browsers.

With my current setup (Mosaic-l10n2.6, Apache 0.6) this doesn't quite work,
as unknown.html is returned when accept-language is set to "en". Putting
a bogus language first has the desired effect, but this is obviously a 
kludge.
http://vancouver-webpages.com/multilingual/samples8.var, samples8.var.txt 

Alternatively, I could just add all the other languages to the english 
one, so that I have:

URI: english-ca.html
Content-type: text/html
Content-language: en-CA,en-US,en-GB,de,fr-FR

which would work, but would mean I would have to clutter up the "English" 
page with instructions about setting up Accept-language.

Andrew Daviel

andrew@vancouver-webpages.com 
http://vancouver-webpages.com  : home of searchBC

Received on Sunday, 14 July 1996 12:11:35 UTC