meeting?
John Jannotti
jj at lcs.mit.edu
Mon Mar 22 17:09:15 EST 1999
Do you guys want to meet tomorrow? I've worked most of the old ipb
stuff into the new framework, but have a few questions.
Suppose I have:
faction 0: dumpreader - push
faction 1: wire - push
faction 2: queue - push-to-pull
faction 3: wire - pull
faction 4: print - pull
How is the print faction supposed to work? Should it just constantly
return true from ready() and thereby poll the queue through the second
wire? I think the key here is just that we don't have anything for
directly supporting faster outputs than inputs in the current driver
without polling. I also think such a thing might be easier to add if
we went back to the "1-call-depth" driver model (since state is more
explicit then), but I'm not completely sure of that.
What is out stance on when factions are connected to different
input/output number? I know we've talked about only using different
inputs for different semantic meanings, such as within ARP, but I'd
like to connect each input of the etherswitch to a different input.
Does that sound reasonable?
Do 10/100 ethernet switches have buffers for going from 100 down to
10?
jj
More information about the click
mailing list