On Mon, 29 Jan 2001, Joshua Slive wrote: > On Mon, 29 Jan 2001, Koen Holtman wrote: > > Wait a minute: is the problem that these many people send, say > > > > Accept-language: fr > > > > when they in fact speak en+fr and prefer en, or is the problem that their > > browsers do not send any accept-language at all? > > > > The problem is people who send > Accept-language: de > or > Accept-language: en-gb > when their "true" preferences are "de,en" or "en-gb,en". Since the server > has no "de" or "en-gb" variant, they get the "No Acceptable Variant" page. > > People sending no accept-language at all isn't a problem becuase, I > believe, the LanguagePriority directive will then do the necessary magic > and send the english page. The problem is people who lie about their > language preferences becuase they don't understand how to configure their > browsers. > > I was willing to just ignore such people and hope they fix their > configuration, but the problem seems to be more widespread than I > expected.
Unless I've missed something (quite possible, this late at night!), the solution seems to be adding a directive to specify a "language of last resort", which will always be tried before sending the NAV error? As pointed out earlier, this could be easily tailored by mirrors - e.g. apache.jp could mirror the whole of apache.org's content verbatim, but have a different setting for this variable - quite a good approximation to what people would expect, unless they specify preferences we can satisfy. Alternative approach: create a CGI as the ErrorDocument for this error, which will display a delayed redirect to the appropriate language, along with a message explaining what's going on to the user?? James.
