[conspire] tmpreaper and use of /tmp as scratch space

[Rick Moen]

Quoting Ivan Sergio Borgonovo (mail at webthatworks.it):

> meh... I took the habit of mounting stuff with noatime.
> This setup became much more popular with SSD.

Yes, indeed.  

FYI, Debian's tmpreaper implmentation, for one, can be caused to use
mtime rather than atime, if you include the --mtime option in any
invocations.  (I figured getting into that, upthread, was too much
detail.)

OTOH, if you have /tmp mounted from tmpfs, which would be a very
reasonable thing to do on an SSD-based system, AFAIK there'd be no
compelling reason to mount w/the noatime mount option _there_. 
(I mean, unless spillover to swap occurs, it's just _RAM_.  Updating
atime isn't going to wear it out.  ;->  )

> And I just noticed that tmpreaper is not installed by default.
> Has this task been delegated to systemd?

Well, I've not done a from-scratch installation from Official Debian
media of Debian 9 'Stretch', though I installed Debian 8 'Jessie' and
later upgraded it to Stretch.[1]  (systemd was present only until I
could replace it with OpenRC.)  On that system, tmpreaper was indeed not
installed by default.  

That system has only an interim implementation of OpenRC.  I might try a
fresh Stretch installation, to see what that's like, but I'm actually 
more interested in the recent Devuan installers, in part because the 
OpenRC implementation there is a good one (while I'm not sure about the
state of Debian's).


[1] This was an experiment, a couple of years back, so that I could
write http://linuxmafia.com/faq/Debian/openrc-conversion.html using hard
data collected from such a VM-based system.