[conspire] verifying I'm doing this correctly - fdisk, mkfs.ext3...

[Don Marti]

begin Rick Moen quotation of Mon, Mar 29, 2010 at 07:13:17PM -0700:

> Overhead of atime updating isn't[1] all _that_ big a problem except
> (arguably) when running on battery power -- though one can usefully
> experiment with doing "noatime" and "nodiratime" on parts of the tree,
> and it does have significant benefits on my server.  (Some places in the
> file tree, the atime time stamp doesn't matter at all.)  

And some people are adamant about turning on noatime
everywhere.

If I write a program, it has to work on my
power-tweaked laptop, which has / mounted noatime.
And if I share a program with the world, I have to
assume that some of the users are going to be running
with noatime, so I can't really rely on atime.

For example, I wrote a simple music player that stays
out of the way:

  http://gitorious.org/hemp

Uses gstreamer plugins, so will play anything you have
the codecs for, including the audio tracks of your
music video collection.  It's now at Works For Me(tm)
stage, although the m3u playlist parsing needs work.

If I used atime to play less-recently-heard stuff
first, it would break in noatime environments, so I
have to use a separate playlist file.  

And if people who write or maintain software are
assuming that noatime might be set, so the software
has to work without it, then you might as well run
noatime, unless you need to support some legacy
thing that still has no noatime weenies as users.

Or am I missing something?

-- 
Don Marti                                 +1 510-332-1587 mobile
http://zgp.org/~dmarti/
dmarti at zgp.org