On Thu, Jul 23, 2020 at 00:01:18 -0500, Derek Martin wrote: > Oops, if I'd seen this message before I sent my last post, I probably > wouldn't have bothered to post it. > > That said, I will take issue with the notion that mbox is a terrible > format: It isn't. It does, however, have usage patterns for which > it is not well suited...
I'd like to point out that I said that mbox was somehow bad only twice in my
email and in both instances there were qualifiers about the workload:
"mbox is a horrible format for any sort of mailbox that sees
modifications other than just appending"
and:
"[updating the status info] may require rewriting the whole mbox
file - which is the big reason mbox is a terrible format :)"
(The second is a bit more implicit about it than I wish, but having to
rewrite the whole file *is* a major flaw with mbox.)
> just like maildir does. If you've read the
> famous comparison of the two on the Courier website, you should note
> that while many of its claims are true, it nevertheless is in no small
> part bunk.
I may have seen it years ago but I didn't remember anything about it. I
looked at it just now, and I have to say that I wouldn't trust the results
for anything today. 17 years ago they *may have* indicated something
useful, but I'm not even convinced of that. The only thing I truly believe
from its final analysis is:
These benchmarks show that a big factor is the underlying hardware
and the operating system.
Which more or less invalidates its other results.
...
> I'll grant you that one or two of the points I made may be outdated,
> depending on what hardware and file system you're using and what
> options you're mounting it with. But mostly... probably not.
There are definitely use cases where mbox is a decent choice. I used to use
maildir for all "regular" mailboxes and mbox for the spam mailbox, but I
both to more efficient formats in Dovecot (sdbox and mdbox, respectively).
These more efficient formats roughly correspond to maildir and mbox, but use
external index files to keep track of status and other often modified data.
This reduces the number of fs operations quite a bit.
Jeff.
--
Intellectuals solve problems; geniuses prevent them
- Albert Einstein
signature.asc
Description: PGP signature
