On Thu, Jun 25, 2015 at 02:21:47AM -0000, John Levine wrote:
> Speaking of case folding, I note that nobody here seems to understand
> UTF-8 and EAI.
That's a unnecessary over-generalization. Some of us have read
enough of the 2003 and 2008 IDNA standards to have been grounded
in the various issues. I've been involved in multiple discussions
on EAI implementation in a popular MTA.
My comment about Cyrillic was an observation of current practice
in major browsers with at least one non-Latin (though still European)
script, there are clearly some pragmatic choices being made at the
user-interface layer that would not be appropriate in most other
contexts.
Thus while MTAs, DNS servers, ... must not case-fold UTF-8,
user-facing software (such as a mobile phone or similar), where
users enter email addresses, might well have enough information to
do something sensible with mixed case input, and may allow users
to correct unwanted normalization (damn you auto-correct).
Of course the story is much simpler for IDNA domain names than it
is for EAI localpart addresses. IDNA (2008) domains use a restricted
character set that does not allow for the existence of two distinct
valid domain names that differ only in case.
The corresponding isssue for local parts is much more difficult,
because email addresses are not expected to satisfy this constraint.
I might hypothetically have an email address of:
Виктор.Духовный@духовный.рф
In which, when entered by a user into an EAI-enabled MUA, the domain
might be casually mapped to lower-case Cyrillic, but the wisdom of
a similar localpart mapping is much more questionable, and might
require user confirmation.
What this might mean for the draft under discussion, is that any
case-conversion might need to be user visible and not implicit,
and be a feature of the address input UI, and not the underlying
PGP key lookup mechanism, once the user settles on an input address.
We can sensibly discuss other approaches, without asserting that
everyone else present is clueless. If I and/or others have the
wrong end of the stick, a more constructive suggestion would work
better. This could be, lower-case lookup for ASCII only, or
could be lower-case lookup only with user confirmation, (feature
of MUA user-interface, not key lookup library), etc.
--
Viktor.
_______________________________________________
dane mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dane
