-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
On Friday, September 4 at 09:49 PM, quoth Chris G:
>>>>> Depending on what/who created the maildir hierarchy you may
>>>>> find it virtually impossible to move directories (which aren't
>>>>> real directories) and mailboxes around.
[snip]
>>> Mail.holidays.2009.sarthe
[snip]
>> I fail to see how this makes renaming or moving the sarthe mailbox
>> "virtually impossible".
>>
> It makes it extremely awkward if you have some mailboxes at levels
> part of the way down the hierarchy, it also make the names painfully
> long.
Okay, we clearly have different understandings of the term "virtually
impossible". I read it as "cannot be even theoretically accomplished,
except maybe in some unusual circumstances", whereas you seem to
define it as merely "extremely awkward". By that metric, all of the
walks in the Monty Python "Ministry of Funny Walks" sketch are
"virtually impossible".
In any event, I disagree about how awkward it is. It requires more
steps to move multiple mailboxes at the same time, yes. But that makes
it more involved, not more awkward. Compiling mutt from source
requires more steps than installing a pre-fab package, but that
doesn't make it awkward.
>>> I note you haven't questioned my comment that maildirs are
>>> difficult to delete safely! :-)
>>
>> They're only difficult to delete "safely" if you're expecting them to
>> get recreated at any moment. And that's only true if you need an
>> algorithm that doesn't impose arbitrary naming restrictions. If you're
>> allowed to do this:
>>
>> mv maildir ..maildir_deletemeXXXXXXXX
>> rm -rf ..maildir_deletemeXXXXXXX
>>
> But those caveats apply in *exactly* the same way as your criteria for
> maildir being 'safe' for access by multiple applications!
*My* criteria? Perhaps I missed an email here; at what point did I
establish these criteria?
In any event, I was wrong about the renaming business. There's an
inherent race condition between knowing the contents of a maildir and
choosing to delete it. An atomic rename doesn't help, because there's
still a race between knowing the contents and renaming it. This can
only be done safely if there is some external locking mechanism to tie
those two actions together. I don't think this is a big problem,
but... <shrug>
~Kyle
- --
Our lives begin to end the day we become silent about things that
matter.
-- Martin Luther King Jr.
-----BEGIN PGP SIGNATURE-----
Comment: Thank you for using encryption!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=r4XN
-----END PGP SIGNATURE-----
