Re: IDNs in IE and Google

Michel Suignard wrote:
> Concerning IRI, it is not a matter of 'preference'. If you present
> something like a URI containing a host name presented in non ASCII
> repertoire, you are in fact using an illegal URI per RFC2396 definition.

While this is true, it only imposes constraints on the HTML authors -
not on the Web browser. In particular, B.2.1 of HTML 4.01 explains
that Web browsers should expect authors to violate the constraint,
and gives a suggestion on how to deal with this violation. This
suggestion clearly predates IDNA, so dealing with this violation
using IDNA for the host name part is more sensible.

It is (unfortunately) quite common that HTML authors do not follow
the relevent HTML recommmendations. Some authors even believe that
unlimited extensibility is one of the guiding principles of HTML.

In fact, MSIE 6 does process URLs with non-ASCII characters - even
though not in the suggested way (but instead, apparently by sending
UTF-8 directly to the wire). Changing it to perform IDNA on the
host part (and leaving everything else as-is) would not make it less
standards-conforming, and give users added value.

Regards,
Martin