In message <20672.1131936697@splode.eterna.com.au>, matthew green writes:
>
>   >    IMO removing tmpfs auto-sizing feature is a way to go.
>   >    ie. mandate -s option of mount_tmpfs and trust it.
>   > 
>   > 
>   > i want tmpfs to resize based on how much ram i have, or swap, or
>   > some cmobo, by default.  having to change static configuration
>   > parameters when i simply add memory is a concept that should be
>   > dead in this millennium...
>   
>   i don't think that the appropriate size of the filesystem is
>   directly related to the total amount of memory on the system.
>   it's rather related to the purpose of the filesystem.
>   at least, it can't be mechanically calculated from the amount of memory.
>
>
>i'd hope that one day tmpfs on /tmp is the default config for
>a freshly installed netbsd system.  i don't think that it should
>require an explicit size and should just vary itself based on 
>the amount of ram/swap/something.
>
>i'm not saying i think it should use ALL ram/swap, but it seems
>the wrong thing to do to force size to be explicitly chosen.
>
Given the amount of (justifiable!) complaining we already see about how 
hard it is to get VM tuning correct, I don't think this is a good path 
to follow -- it adds one more element of unpredictability to system 
behavior.  At most, I think that sysinst could look at RAM and swap 
size and suggest that as the default -s option for /tmp.

		--Steven M. Bellovin, http://www.cs.columbia.edu/~smb