Issues lists and the "preprocessing" topic

[Andrew Sullivan]

On Wed, Aug 20, 2008 at 10:11:00AM -0700, Lisa Dusseault wrote:

> sophisticated about URIs than browsers are.  Text editors, WYSIWYG  
> editors, calendar clients, IM clients, presentation authoring  
> software, file readers (like PDF readers) -- these all may present  
> URIs that users can click on and allow users to enter non-ASCII  
> characters into such URIs. 

This discussion strikes me as an example of the thing I was worrying
about right after Dublin.  In the hope of trying to make clearer what
I was yittering about that time, I'm going to try again.  I think
there may be two cases, and some of us ("at least one" may boil down
just to me!) may sometimes conflate them.

One of these cases is what we might call "narrow IDNA context".  I
think of this as applications using names as part of the resolution
step of performing network operations.  In some cases, those names
include at least one U-label.  It seems to me that addressing this
context can be boiled down to a strict problem of what gets encoded in
the ASCII-compatible encoding, what cases are tricky (and how to deal
with them), and what cases simply won't be allowed at all. I believe
some of the WG participants think this is the entire problem that the
WG should address, and that anything else is something entirely
outside our charter.

Another case is something we might call "wide IDNA context": the way
in which applications use U-labels, what they ought to do with them
when handing them around, indexing them, &c.  This is not analogous to
gethostbyname() and friends, but analogous to how traditional FQDNs,
URIs, &c. are interpreted by network-aware (but not
"network-competent") applications today.  I think that those who
believe the narrow context is the only thing on charter here believe
that all of these issues are just problems high up in the application
space, and that they're none of our business.

I think that IDNA2003 tried very hard to cleave to the narrow context,
and left most of the issues to the application's space.  I think that
the attempt to unhook from a specific version of Unicode can
nevertheless be made to fit in the narrow context.  I have a sneaking
suspicion that attempts to ensure stability with IDNA2003, and
especially any other local mapping, is really a wide context problem.
I'm undecided about whether wide context problems are on-charter in
this WG.

If I have completely misunderstood everything (it's not the first
time), I'd appreciate the relevant poke in the correct direction.

Best,

A

-- 
Andrew Sullivan
ajs at commandprompt.com
+1 503 667 4564 x104
http://www.commandprompt.com/