Re: [IEE] WG Review: Email Address Internationalization (eai)
Yangwoo Ko <newcat <at> icu.ac.kr>
2006-03-03 09:18:53 GMT
Dear Dave Crocker,
Thanks for good points. Please find my replies embedded below.
Dave Crocker wrote:
> > Dicussion between Dave and John, quoted above, can be understood as how
>> significant it will be? I'd rather interpret the "significance" in
>> somewhat different aspect.
> I think you have describe the question correctly.
> I think that a change in the form of email addresses that can cause
> delivery to fail should be considered significant.
(Off-topic) Are we "changing the form" of email addresses? As a
speaker, it is always very hard to grasp various meanings and usages of
I think we are just "adding more forms" thanks to the nice property of
(On) Delivery failure is not always bad. Through a failure notice, a
user is able to see that
he/she should use alternative addresses if available. Though I don't
think this type of human
assisted address alternation is the best engineering solution, it is
still one of valuable
options we can take.
> It can/will partition the functional infrastructure.
>> As we assume, it seems to be relatively easy to do IMA only between
>> agreed parties. The real devil lies at the boundary.
> Yes. Exactly correct.
>> However good news is that, as a design team of this extention, we
>> have a knob that controls how significantly we want to change the
>> infrastructure. For example, in one extreme, we can just drop any
>> mail that tries to cross the boundary. On the other extreme, every
>> part of the Internet infrastructure is required to be upgraded
>> somehow to support IMA so perfectly that users cannot even recognize
>> that there is such boundary. In order to make IMA more usable, we
>> want to deviate from the former extreme as much as possible. In order
>> to make IMA be more practical, we cannot but deviate from the other
> How does a design team control the operational behavior of boundary
> MTAs? The experience with gatewaying between independent mail
> services is that it is a challenge, at best.
I don't see your point. If a standard cannot define the behavior
of "compliant" components, then what is that standard for?
Or, are you saying that it is impossible to standardize the
behavior of gateways because the complexity of gatewaying
is simply beyond the human-being's brain power?
> Further, the experience with the use of "private" IP Addresses within
> networks is they leak out into the public Internet.
Yes. They will leak regardless of how hard we try to
design the right thing. However, we have defined private IP address
range and users are using it everyday. Because it is very useful to
solve many real issues.
No one can say that I18N addresses will not leak at all. Users who
are not using EAI-enabled components will get hurt from time
to time. It seems impossible (or at least impractically hard) to avoid
this type of uncomfort completely. If IETF as a community
has a rough consensus that any form of leak with any frequency is
significant and any new idea leading to that significant change should
be dealt by IRTF, then let EAI be discussed at IRTF.
> In the last 25 years, there have been somewhere between 0 and few
> efforts to affect basic deliverability. (DSN is the possible
> exception and it's adoption history provides quite an education.)