[MlMt] Delete Autocompleted E-Mail Address From To Field

[Benny Kjær Nielsen]

On 13 Mar 2013, at 8:49, Vincent Noel wrote:

> On 12 Mar 2013, at 22:54, Benny Kjær Nielsen wrote:
>
>> On 12 Mar 2013, at 17:34, Bill Cole wrote:
>>
>>> Rather than requiring an explicit marking of addresses, I'd rather 
>>> see MM construct & sort the list more intelligently. One improvement 
>>> would be to canonicalize & de-dupe addresses so that only the most 
>>> frequently used version of functionally identical addresses is 
>>> offered. For example, if I type my boss' first name into the To 
>>> field, I get 7 choices in the pop-up, all of which have one of his 
>>> addresses (his name is not very common) and 5 of which share 
>>> functionally identical email addresses, with variations in the "real 
>>> name" part and address capitalizations. I don't really care much 
>>> which of those 5 I use, but it would be better to have 3 choices 
>>> rather than 7, with the first one being the most commonly used form 
>>> of the most commonly used address.
>
> I'm seeing that too, a lot.
> Here is a pathological example:
> http://db.tt/iZjaPDxs
>
> I would be very happy if there was a way to de-dup the list.

That's a great example. It also shows how non-trivial this problem is. 
I've just taken a quick look at the code and I don't think there are any 
bugs, but there is certainly a need for some aggressive heuristic for 
removing duplicates. The problem is that this is far from trivial.

I've added that addresses like 'foo at bar.com <foo at bar.com>' are treated 
like 'foo at bar.com'. This also means excluding them if a named variant 
exists ('Foo Bar <foo at bar.com>'). That one was easy (if no one has any 
complaints).

For other changes I would like to follow this principle for new 
messages: If possible, the recipient of a message should be the one 
deciding which address format to use (just like for replies).

Currently I only look at sent messages and this is kind of wrong. The 
to/cc headers of sent messages are good hints with regard to who you 
write to, but it is the from header of received messages that shows how 
addresses should be formatted. The from header may even be better than 
the name used in the Address Book (again, just like for replies).

The big problem is how to decide when two names for the same email 
address are sufficiently similar to warrant the removal of one of them. 
Maybe I should just start out with the opposite extreme of the current 
behavior: An option to enforce at most one completion suggestion for a 
given email address.

-- 
Benny