>>>>> "cws" == Carl Wilhelm Soderstrom <chrome@real-time.com> writes:

   cws> it's been hammered over before on the list, and I think the
   cws> upshot was that you shouldn't worry about it.

must have been a different list.

   cws> it shouldn't swap much anyway.

I have 32M RAM and 19MB swap used right now.  I'm running a
recently-booted emacs in an xterm.

The thing is, doing builds over NFS or SlimSCSI, with noatime and
nodevmtime mount options, and no swap, the CF card won't be written
while loading and running applications, only while saving files or
emails or whatever.  A typical swapless NetBSD box could do very
little disk writing, and this is similar to other uses of FLASH like
digital cameras, MP3 players, and firmware upgradeability.

With swap, the usage pattern is completely different from the
applications that FLASH designers were forced to consider.  so I think
the hand-waving argument against CF FLASH is actually pretty
convincing.  

   cws> (1M cycles comes to mind). this will probably take a
   cws> surprisingly long time to accomplish tho, on an OS which
   cws> doesn't swap unless it *really* needs to.

What would be more interesting than the hand-waving arguments we've
both proposed is someone who has actually done it, to know how much
swap space that 'top' or 'swapctl' show used and how long they've been
doing it, and whether or not the FLASH has gotten slower over a year
or two.  This is probably an affordable experiment now, although
keeping someone's attention for a whole year might be a challenge:

   cws> you ought to get a small laptop with more memory and a
   cws> spinning disk (I'm reasonably happy with my Dell X200).

great.  here we go.

-- 
This may sound corny, but have you given any thought to your immortal
soul?
                -- Kathleen E. Carberry