Esszett, Final Sigma, ZWJ and ZWNJ

Paul Hoffman phoffman at imc.org
Tue Feb 24 02:06:47 CET 2009


At 3:17 PM -0500 2/23/09, Vint Cerf wrote:
>For clarity's sake, can we look at the "bits on the wire" issue for a moment?

Maybe even more than a moment! :-)

>Would the addition of any new character to the allowed set constitute changing bits on the wire?

Adding any new character is not of concern; it is certainly what we all have in mind here. Removing some old characters is also not of concern, and is also what we all have in mind here.

Changing an "old" character that IDNA2003 mapped to another character into IDNA2008 PVALID (which is, of course, unmapped) is changing bits-on-the-wire.

>Is it this change that captures your concern for "bits on the wire" or have I not understood the point?

There is an addition, however: an IDNA2003 label that had both an Esszett and another non-ASCII character would already be encoded with Punycode; under IDNA2008, it would still be in Punycode, but it would have a different value. This means that some labels will go from non-Punycode to Punycode, while others will go from one Punycode string to a different Punycode string.


More information about the Idna-update mailing list