On Mon, May 15, 2006 at 10:50:03AM -0400, David Rosenstrauch wrote:
> Hmmm.  I thought I remembered reading somewhere that the way tmpfs 
> worked was that if it didn't have sufficient space in RAM, then it would 
> use swap space for the rest.  So if I specify 1GB, I will have access to 
> that amount of space in /tmp, though not all of it will be in RAM at any 
> given time.  Seems like a reasonable way to assure sufficient /tmp space 
> to me, and would address the k3b problems.

Yep, you're right about that.  But it's still not as solid as using
real, non-swap diskspace.  If a user only has 256mb RAM and follows the
age-old convention of dedicating 2*RAM to their swap, then they'd only
have 128mb+512mb = 640mb of actual tmp space, and that's with their swap
partition completely tapped.  If the user installing Arch KNEW that we
used tmpfs, then s/he could account for that by creating a bigger swap
partition.

It's still there for the pros, but the default=off seemed simpler for
everyone.

> Also, if you don't have /tmp assigned to tmpfs space, won't you have old 
> temp files hanging around that you don't want?  If you assign /tmp to 
> tmpfs space, then you're guaranteed that it will be wiped clean on next 
> reboot.

/tmp gets cleaned in rc.sysinit at boot-up, along with some other areas
like /var/lock and /var/run.

> Anyway, I was just wondering what the reason was for the change, which 
> you provided.  Thanks for the FYI.

No problemo.


- J


_______________________________________________
arch mailing list
[email protected]
http://www.archlinux.org/mailman/listinfo/arch

Reply via email to