On Wed, Jun 08, 2011 at 05:54:20PM +0200, Jörg-Volker Peetz wrote: > In addition to my bug report I would like to suggest to use something like > /run/man instead of /var/cache/man, i.e., keeping the whole man page cache in > RAM.
I don't think this would be appropriate in general. If you want to symlink your /var/cache/man into a tmpfs such as /run, or mount a tmpfs there directly, you're welcome to do so. > On my (personal use) systems the size of /var/cache/man is below 3 MB. > I have no idea how big this directory grows if every package is installed. Not all that much bigger. I intentionally have lots of manual-page-heavy packages installed for testing (though certainly not all), very likely more than most, and /var/cache/man is only 12MB here. Any system with enough space for that many packages can afford that, and in any case RAM is more expensive than hard disk space. > This would have the advantage of avoiding several hard disk write and also > read > accesses. A handful of writes, but the net effect would be very many more reads. Reboots are not *that* infrequent. > If this suggestion will be followed, the script /etc/cron.daily/man-db should > also be started after any system start. And that would kill boot performance, because it would have to rescan every file in /usr/share/man just to make apropos work again. Even on an SSD that takes a while - two and a half minutes here. You don't buy an SSD so that your system can spend two and a half minutes reading /usr/share/man at boot time ... There are certainly ways to optimise mandb, which I do plan to work on for a variety of reasons, but I'm afraid I don't think this is a good one at all. Cheers, -- Colin Watson [[email protected]] -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
