Chris Haynes wrote:
The human language used in a Web page is purely determined
by the author of that page. Nothing the browser or "3" knows
or says (normally) affects that. So whoever writes that
page has complete control over its content.
No, that's not the idea of "content negotiation" depending on
the language preferences sent by a browser in its GET request.
You can have a single URL, and depending on the preferences
of the browser the http-server can send different replies.
But there's another problem, much worse than all I had already
mentioned: In my example (1 = policy author, 2 = user, 3 = MSA,
and 4 = MX) only "1" and "4" necessarily know SPF at all.
"3" can be a 3rd party, which has no idea what the hell SPF is.
E.g. "1" = pobox / "4" = AOL (or vice versa), and the typical
problem in your scenario "legitimate error" is then a user of
a pobox (resp. AOL) address sending mail via a 3rd partv to
AOL (resp. pobox). The 3rd party ("3") has no idea what's the
problem between "1" and "4", all they see is the error message
of "4", and pray they forward it literally in a bounce to "2".
BTW, that's also the reason why there are no correct accents
in the French explanation: The error message is inserted into
a bounce by a 3rd party, and this 3rd party has its own ideas
about the format / language / charset of its bounces.
it is "1" who makes the decision on the language to be used.
Yes, and if they are smart they offer "content negotiation"
depending on the laguage preferences of their users ("2").
Of course they can also offer a default page with links to
different languages. But whatever they do, it's not the job
of "3" to create a multipart/mixed bounce with a text/html
part fetched from "1".
"3" is a MSA, it uses ports 25, 53, and maybe 587, but not 80.
Bye, Frank