Leonard Erickson wrote to Rodrigo Cesar Banhara on 10-19-1999

LE> One of the simpler aspects of the difference is that a synchronous
LE> link doesn't *have* start bits and stop bits. Thus 8-bit, no parity
LE> data moves 1.25 (10/8) times as fast over a synchronous link. 

AH, memories of the good old days of writing bisync drivers for a
pre-Grid laptop!

LE> You do pay for this. Anytime a new byte isn't ready to be sent in
LE> time, you have to add a "dummy" byte to maintain synch. And usually,
LE> you dedicate one byte value to this, and have to use some sort of
LE> escape char to actually send that value. Even so, synch is still
LE> faster than async.

Depends on the sync protocol supported.  Bisync (in the 8251 and 8273,
at least) included a length count after the sys-syn pair which then
allowed the syn character to pass as data.

LE> synch links tend to be *inherently* bi-directional also. 
 
LE> The main reason I'm interested in synch links is because I've got
LE> a pair of CSU/DSU units that can do 56k if connected to a sync
LE> interface, and only 19.2k if hooked to an async one.  

Hadn't thought of this in years.  What I really remember was the
hassle of getting the protocol spec out of IBM to be able to write the
drivers.  In the days of 300 baud acoustic modems this allowed up to
9600 with direct connection - INS800 processor couldn't keep up with
anything faster.  Now we complain if we get 'only' 33.6!

Will Honea 
--- Maximus/2 2.02