microsoft claims
David R. Conrad
david.conrad at nominum.com
Sat Oct 21 23:40:28 UTC 2000
Hi,
> > > > 3. bind does not support utf-8 character encoding.
> > This is false.
>Is it legal to use utf-8 ? Seperate from is it supported or not.
Well, I don't think the police will hunt you down should you put a UTF-8
string in your domain name server... :-)
More seriously, RFCs 952 and 1123 define that a "hostname" should consist
only of A-Z, 0-9, and "-". Clearly, most UTF-8 values violate this
definition. However, a domain name can consist of any value and there is
_very_ strong pressure (particularly out of Asia) to internationalization
of the DNS. One of the approaches is to use the approach Microsoft has
taken, namely simply putting UTF-8 on the wire. Unfortunately, this
approach has several flaws:
a) it breaks software (like MTAs, MUAs, etc.) that "knows" host names can
only be A-Z, 0-9, and "-", sometimes in particularly unpleasant ways.
b) as it is variably multi-byte, the use of UTF-8 can significantly limit
the number of glyphs one can use in a name for some languages
c) Microsoft did it, therefore it must be inherently evil (well, not
really, but that seems to be the impression I get from the IETF sometimes)
Another twist is that there are many known bugs in UTF-8 implementations
(Microsoft's encoding and decoding in particular) that make trying to
figure out what a UTF-8 value is supposed to be a bit "challenging".
On the other hand, using UTF-8 works (most of the time) with modern
browsers, so people can make use of UTF-8 strings in the DNS today.
However, in the context of the original post, even the folks in Microsoft
who wrote their DNS software have publicly stated at the Pittsburgh IETF
that they consider their approach of "just send UTF-8" a bad idea, so I
find it a bit weird that they are claiming this is a good thing. Marketing
!= Reality, I guess...
Rgds,
-drc
More information about the bind-users
mailing list