Re: [EAI] NFC/NFD (Re: Test - driven schedule (Re: Rechartering))

[Harald Alvestrand <[e-mail]>]

Arnt Gulbrandsen wrote:
> John C Klensin writes:
>> --On Friday, July 24, 2009 09:22 +0200 Harald Alvestrand
>> <[e-mail]> wrote:
>>> The more dangerous brokenness is the business card data entry 
>>> situation, where people enter an address from a business card and 
>>> get a "no such mailbox" response, without there being any visual 
>>> difference between the two strings.
>>>
>>>  Requiring normalization at the server (without placing a 
>>> requirement  on what one normalizes to) would at least remove this 
>>> class of  issue.
>>
>> Exactly.
>
> I wonder if that's a bit of a fallacy... anyone who expects humans to 
> read and reenter 42-69 differently from 42‒69 has a problem. IMO 
> that's a perfectly natural mistake, it involves a character often used 
> in today's email addresses, and I've never heard anyone use this 
> particular example before so I doubt it would be implemented quickly.
and in this case, NFC wouldn't help either....
>
> Alternative proposal: Extend VRFY to optionally return a fuzzily 
> correctified version of an email address. If VRFY is disabled or 
> doesn't return an address, correction isn't available for that 
> address. I think this is better because it permits the fuzzy fuzziness 
> to be invoked when the address is entered into e.g. an address 
> database, instead of when a message is transmitted. As a bonus it's 
> 99% independent of EAI.
Fuzzy matching is a very interesting subject, and I'd argue that it's 
100% orthogonal to EAI - I still remember the mail I got as a new 
postmaster after a new fuzzy-match algorithm had rerouted a misspelling 
of one man's wife's mail to his mailbox - "how did the system know we 
are married?".

I'd like to punt this issue to the Single Global Directory, but that's 
just reducing it to a previously unsolved problem.....

                      Harald