Powder and IDNAbis

I had an action item to bring the Powder IDN issue to the IDNA Working
Group. Below is a summary of the resulting discussion.

IDNAbis defines A- (ASCII only) and U- (Unicode) labels. In IDNAbis it
is planned to gurantee 100% reversibility / round-tripping for the
conversion between A- and an U-labels. This means that mapping of
certain characters (e.g. "Eszett" to "ss") as defined in IDNA2003 are
taken out of the conversion between A- and U-labels.

However, "taken out" does not mean that these mappings will never happen
anymore. Esp. Mark has presented earlier a separate document which
describes the necessary processing. It is currently unclear whether we
will see another kind of label, by somebody called "I" label, in
addition to "A" and "U" and being outside of the IDNAbis protocol.

What does this mean for Powder? I see the following choices (including
reasons for making a particular choice):

1) if they want 100% reversibility between A- and U- lables, they can
refer "just" to IDNAbis in their spec. Currently this would be a
placeholder, since there is no RFC number yet for IDNAbis. Drawback:
they will loose the possibility of usng characters like "Eszett".
2) if they want to maintain using "Eszett", they can refer to IDNA2003
and keep the mappings. Drawback: IDNA2003 will be obsoleted at some point.
3) if they want to maintain using "Eszett", they can refer to IDNAbis
and the separate processing document mentioned above. Drawback: that
document doesn't have any official status yet, neither within the IETF
nor in a different organization.

Each choice is an 80/20 solution. 1) gives higher importance to
reversibility compared to deployed technology. 2) and 3) are the other
way round. As far as I can tell there is no silver bullet here, and I
think we should explain this to the Powder WG and see how they think /

Some additional remarks: nobody in the IDNA Working Group asked "They
does Powder need this?", so I did not even try to explain that. Also I
did not mention case folding which we discussed at last weeks call,
since I don't thin that the Powder folks or others expect that
"someDomain.example.com" is different from "somedomain.example.com", or
that somebody is trying to make them different in IDNAbis.

The discussion took place via mail, see the thread at



Received on Wednesday, 28 May 2008 10:52:24 UTC