V.32 and V.34 modems are full duplex. Modulo hardware implementation of
course. It is rumoured for example that USR Sportsters cannot handle full
28.8kps bi-directionally because it simply is not fast enough. The USR
Couriers can though. Most modern implementations of quality V.34 modems
should not have any problems on good lines.
>That leaves the poor peformance of a 28.8 dialup compared to half that
>of a 56k line unexplained. Modem turnaround time is the explanation I
>have often seen given.
Perhaps they are mis-describing the high round trip time.
I.e. Your round round trip time on a sync 56kbps circuit is under 40ms.
Compared to maybe 150ms for an async 28kps line.
The delay is due to two things. First of course is the difference in speed.
The bits are clocked through twice as fast. And for sync we only transmit 8
bits per char plus a bit of framing vs 10 bits for async.
Second modems tend to have a slightly higher latency for transmitting data
because they buffer it and use V.42 to compress. This works well to increase
throughput. But at the expense of increasing latency. CSU/DSU's on the other
hand (usually) just clock sync bits in on one side and out the other side
with one character hold time (i.e. the first bit on a received octect
should be starting out the far side one bit time after the last bit was
received).
Because you actually end up with perceived bit rates quite high by utilizing
V.42 it is left in. But it actually is probably a major cause of latency.
If I remember correctly typical V.42 implementations will buffer around 64
to 100 bytes of data trying to compress. At 115kbps this could amount to
around 10 ms of delay (accounting for perhaps an additional 20ms of round trip
time delay).
-- Stuart Lynne <sl@wimsey.com> 604-933-1000 <http://www.wimsey.com> PGP Fingerprint: 28 E2 A0 15 99 62 9A 00 88 EC A3 EE 2D 1C 15 68