Re: More babbling about speed

[Moon@MIT-AI]

I changed QFILE not to update the who-line for every packet on read
and write; instead the who-line is updated when the process goes blocked.
This seems to make reading in CHSNCP 10 seconds or so faster, although
there's enough variance in the times that it's hard to be accurate.
The who-line still updates every 2 or 3 seconds which seems quite adequate.