<<< Date Index >>>     <<< Thread Index >>>

Re: Why isn't it part of mutt "proper"



Hello Bennett,

* Bennett Todd <bet@xxxxxxxxx> [040818 19:56]:
> A couple of wee-teensy wibbles: my build didn't end up getting the
> updates to the docs in manual.txt;

  6.3.98.  header_cache

  Type: path
  Default: ""

  Path to the header cache file. If unset no cache will be used.
  Otherwise the cache will be enabled for Maildir and IMAP mailboxes.

  6.3.99.  maildir_header_cache_verify

  Type: boolean
  Default: yes

  Check for Maildir unaware programs other than mutt having modified
  maildir files when the header cache is in use.  This incurs one
  stat(2) per message every time the folder is opened.

  6.3.100.  header_cache_pagesize

  Type: number
  Default: 16384

  Change the maildir header cache database page size.  Too large or too
  small of a page size for the common header can waste space, memory
  effectiveness, or CPU time.  You can use the db_dump utility to
  determine the optimal page size.

> and the docs on your web page don't quite match the names of the
> variables, but a strings on the binary | grep cache straightened that
> out.

The docs on my website are for the gentoo version until this day. They
fail to build a manual.txt and the gentoo website came to my site and
looked at the top of my page to find out that I changed the variable
names so I added the manual.txt for the Gentoo section (but I wrote a
few sentences about this).

> One question: the page size tuning (header_cache_pagesize according
> to strings on my binary, maildir_cache_page_size in the docs on your
> web page) I've left default, and I don't have a db_dump on my system
> to check with. I presume that'd be an option if I used Berkeley DB,
> which I don't, don't even have it installed, and don't intend to.
> Any way to check whatever this is with gdbm? Or any guidelines,
> however rude, to the tuning tradeoff involved?

The default is more or less the best you can get. Read: 32k is the
optimum (this is what I messured on Sparc/Solaris (8k pagesize) and
Linux (4k pagesize). I used 16k as default because it is the maximum for
db4 and if I set a higher default db4 falls back on 1k or 2k (I don't
remember), which is signficant slower.

        Thomas