> 
> Well, I thought that the new serial drivers would allow me to complete a
> sup without errors.
> 
> No such luck.
> 
> After about 5 '100 hz hours' - 8 hours or so - my sup process complained
> network connection reset by peer; write error.  And then aborted.
> 
> This is exactly the same symptoms as I had before, which I ascribed to
> lousy serial drivers.

Check the logs. Particularly /var/log/info. If it's full of
zs0a fifo overruns or zs0a ring overruns, you've got serial port
problems. Actually, with the new drivers, these errors would mean
something's camping out with interrupts off for a while.

If you see no complaints about serial errors, you probably have a problem
which crops up from time to time on current-users. Some of the ISP's
have problems with certain data patterns in the sup file, and choke
on the transmission. 

> Is it just me, or is SUP too sensitive?  Couldn't it be stateless, and
> restart after such an error?
> 
> Upgrades of doc work - but allsrc crashes after between 4 an 8 hours.
> Sometimes tantalisingly close to finishing.
> 
> My next strategy is going to be to suscribe to several smaller collections
> rather than allsrc, but does anybody know the origin of the fundamental
> problem here?

SUP'ing smaller bits sounds good. I'm not sure what else to advise.

Good luck,

Bill