[MlMt] Memory usage

[Bill Cole]

I'll jump in here, since Benny said he would be gone for the weekend. As 
always, only he is a definitive source of info...


On 2021-05-14 at 15:53:52 UTC-0400 (Fri, 14 May 2021 21:53:52 +0200)
Patrik Fältström via mailmate <mailmate at lists.freron.com>
is rumored to have said:

> On 14 May 2021, at 20:05, Benny Kjær Nielsen wrote:
>
>> Any strategies for using less memory? [too many virtual mailboxes]
>>
>> I might regret this if there turns out to be unintended side effects, 
>> but you could possibly delete some of the database index files such 
>> as those for the “received” header. Some are more important than 
>> others for the internal workings of MailMate while others are only 
>> needed if searched explicitly.

This is a disconcerting suggestion. Like an aircraft designer suggesting 
that maybe the tail rudder and flaps are not really needed and can be 
ditched to save weight. Maybe...


> Aha, ok, I have not looked at the size of the database files.
[...]
> -rw-------  1 paf  staff   724M May 14 21:42 #quoted#lc.cache
> -rw-------  1 paf  staff   724M May 14 21:42 #quoted.cache
> -rw-------  1 paf  staff   2.1G May 14 21:42 #unquoted#lc.cache
> -rw-------  1 paf  staff   2.1G May 14 21:42 #unquoted.cache
> -rw-------  1 paf  staff   2.1G May 14 21:42 received.cache
>
> The received is indeed a large file. Is any search based on that?


To the best of my knowledge, that file will only be needed if you try a 
search (or have a smart mailbox) that uses the Received header(s) of a 
message. The Received headers form an audit trail of the transit of a 
message from the author to final delivery, with an additional Received 
header added each time the message is handed off from one server (or 
process) to another. If you don't do mail forensics on a regular basis, 
you are likely to never miss its index files. Because virtually all 
messages except Sent or Drafts will have 2 Received headers, deleting 
the 3 files that make up the Received index (.cache, .plist, and 
.offsets) will only delete them temporarily, because any new message 
will cause the creation of new files for what MM considers a previously 
unknown header.

Note that the "#quoted" and "#unquoted" files are for tokens in the 
quoted and unquoted text of messages, with the "#lc" versions being for 
the lower-cased versions of those tokens. That's why they are huge, and 
also why you really should not experiment with deleting them to save 
disk and/or RAM space. Despite Benny's suggestion, I would be extremely 
averse to removing any of the index files with '#' in them, as those are 
for logical constructs of MM, not actual headers.

--
Bill Cole
bill at scconsult.com or billcole at apache.org
(AKA @grumpybozo and many *@billmail.scconsult.com addresses)
Not Currently Available For Hire