Re: Handling "via" addresses

On Thu, Jul 07 2022, Sam Steingold wrote:
>> * Roland Winkler <[email protected]> [2022-07-06 15:15:39 -0500]:
>>
>> On Wed, Jul 06 2022, Sam Steingold wrote:
>>> Subject: [PATCH] Use `mail-header-parse-address' instead of
>>>  `mail-extract-address-components'.
>>
>> There is also `gnus-extract-address-components' that is supposedly
>> faster than `mail-header-parse-address', though less accurate.  Is this
>> an issue?  How are these two functions used elsewhere?
>
> please see https://debbugs.gnu.org/cgi/bugreport.cgi?bug=56422
>
> Lars (who re-wrote gnus and wrote `mail-header-parse-address' writes):
>
>> mail-extract-address-components is a DWIM-ish thing that doesn't have
>> much documented behaviour -- it just tries to make things "pretty" by
>> applying lots of (mostly misguided) heuristics.  So talking about
>> "correct" here isn't er correct.
>
>> If you have an RFC822bis From header, you should use
>> `mail-header-parse-address'.  If you have something that's vaguely like
>> a mail header and want to split it, use `mail-header-parse-address-lax'.

I do not know the history.  I vaguely remember that BBDB v2 contained
some remarks that BBDB up to v2 and mail-extract-address-components have
shared a common history.  Even if mail-extract-address-components
applies mostly misguided heuristics, switching to
mail-header-parse-address can give rise to surprises for some users.
I want to look into this more carefully.