Re: [idn] Re: Roll-out times

[Keith Moore <moore@cs.utk.edu>]

> At 8:21 PM -0400 8/30/00, John C Klensin wrote:
> >Once we do i18n DNS names (directly or indirectly), there will
> >be instant pressure to permit i18n externally-referenced file
> >system names and, of course, email addresses.
> 
> Yes, but that's not what we are discussing here. We're talking about
> the rollout of applications, resolvers, and (if needed) DNS servers
> to handle IDN.

we need to confine the burden to just applications and DNS servers.

the applications have to accept native input and convert to 
on-the-wire ASCII encoded format, and to display ASCII encoded
IDNs in native format.    (this often needs to be done within the
application rather than in the resolver, because the application 
will often need to use the ASCII encoded format internally -
say in RFC 822 headers.  of course the resolver can provide
an API to help the application do the conversion but it can't
be buried in the resolver).

the DNS servers probably have to accept zone files in native format, 
convert them to Unicode, normalize, canonicalize, and (if necessary) 
generate additional RRs for alternate spellings, for the convenience 
of people editing zone files with native tools.


> The question is: if someone wants to associate themselves (such as
> through registration) with an IDN, how long will it take for 90% of
> the *target* users to be able to use that IDN? Non-target users (who
> might take longer to upgrade) are not an issue.

the problem with this kind of analysis is that the target changes with 
each user and with each new mail message (using email as example).  
there are communities with which I can exchange fairly arbitrary MIME 
messages (because they tend to use PCs with graphic displays), and 
there are other communities with which I cannot do so (because their
mail systems still tend to be character-based).  In practice, I have 
to keep track of which community has which capability in order to 
communicate effectively.  

so one might better phrase the question :

"how long until X% of senders who care (presumably, those for which
ASCII is not adequate) can reliably expect to be able to use IDNs 
when communicating with recipients?"  (where "reliably" implies
that it works Y% - say 90 or 95% - of the time)

Y is essentially based on human factors - is a 1 in 20 failure rate
of a recipient to be able to display an IDN sufficient that the user
thinks of IDNs as reliable enough to use?   1 in 30?  1 in 50?

the X figure is the one we're really interested in, it speaks about
how much of the "target market" we're reaching.

> >, what do you think the
> >deployment rate is for
> >
> >* MUAs and MTAs and message stores ?
> 
> The MTA and the message store at the site using the IDN clearly will
> have to be updated before they can accept mail using their IDN. 

I hope not, and I don't see any reason why this should be the case.
(other than IMAP-based message stores that need to be able to search for 
IDNs)

We need to avoid impacting parts of the system that don't need to change.
SMTP can carry encoded IDNs just fine, and to burden SMTP and message
stores with having to support IDNs will delay the deployment of IDNs
by *decades*.

> >* FTP clients & servers ?
> 
> Not a likely early target.

FTP already has I18N mostly defined, its method of encoding filenames is 
likely to be different than the one chosen for IDNs.  But it should be 
fairly trivial for an FTP client to support IDNs.  Unlike email, FTP 
doesn't actually transmit domain names as part of the protocol.

> >* Telnet clients & servers ?
> 
> Unless I'm sorely mistaken, there is no need to update telnet servers
> for any changes in host names. As for telnet clients, they will be
> updated as demand grows.

this is the same as FTP.  telnet clients shouldn't have to change
anything but hostname lookup and perhaps display.

> >I invite you to dispute the figures I'm about to give but...
> >
> >- A large fraction of U**x systems are still running the
> >original Berkeley FTP clients (and Windows 2000 contains, as far
> >as I can tell, a clone of one).  20 years?
> 
> Easy to dispute. Just because they run those today, if IDN becomes at
> all popular, updated freeware clients will appear as popularity
> increases. Replacement FTP clients, many of which already exist and
> have been being distributed in Unix systems for years, will probably
> be updated within a year or so.

In my experience, a lot of commercial UNIX systems still ship with 
antiquated FTP clients and servers (after all, FTP was quite stable
for many years so there was not much reason to change), and a lot 
of UNIX hosts are running releases which are several years old.  
But I don't think this matters much for IDNs, because most of the
burden of IDNs should be on the clients, and UNIX is mostly used
for servers.  Commercial vendors will of course add IDN support 
to their FTP and telnet clients and users who need such functionality
will upgrade.

> >- There are a non-trivial number of people (I don't know if it
> >is anywhere near 10%) who started believing that "cat" was a
> >mail reader back when uucp intersystem mail was introduced into
> >U**x and who haven't changed their minds.   25 years?
> 
> I dispute your statement that they are a non-trivial number. How did
> you arrive at that? This has "red herring" written all over it.
> Clearly, there are still zillions of character-based mail users on
> Unix boxes, but they are most likely running Pine, Elm, or one of the
> Emacs-frobs. All can be updated easily.

using 'cat' as a mail reader is a bit of an exaggeration, but I can
attest that there are still a great many folks running antiquated 
mail readers.  yes, UNIX boxes often have Pine or Elm installed
on them, but they also have 'mail', and many users still use it
rather than Pine or Elm or whatever else might be availble.
(experienced users get tired of learning new interfaces after awhile)

somehow 'mail' hasn't been upgraded to deal with MIME even after all 
these years, so I'm not holding my breath waiting for it to deal with
IDNs either  

Keith

p.s. hmmm, I started working on a MIME-capable 'mail' once upon a time 
(i.e. a modification of the BSD 'mail' program with few additional
bells and whistles other than to support MIME) but got sidetracked...
I should get back to it.  perhaps I could make it deal with IDNs 
while I'm at it.