At 13:34 04/13/99 -0500, Dave VanHorn wrote:
>The break signal is, as far as I can tell, a hack. It induces a framing
>error at the other end, with a byte value of 00, but there is no firm
>definition of how long a break should be.  I've seen "specs" for break of
>100mS, 11 bit times, several seconds.....

i guess the usual minimum is around 1.5 byte times, to be able to
distinguish it from a normal framing error. something like this is built
into some uarts. but in order to be on the safe side, many application
protocols define a longer break -- this doesn't mean that a shorter break
is no break condition, but it doesn't trigger the "break state" of the
application protocol.

one has to realize that even with a simple serial link you have at least
three protocol layers: the physical layer, that is the signals on the
lines, when they are considered high or low, the data link layer, which are
the start, data, parity and stop bit definitions, and the application
layer, which defines stuff such as minimum number of data bits, minimum
break times and such, and which tells you what the lower links have to be
able to provide to be "transparent" to it.

for example, if an application layer protocol defines 7 data bits, all data
link layer which provide 7 or more bits are transparent (with respect to
this requirement). and the application layer doesn't care whether you use
an additional parity bit or better line driver to get down to the max.
allowed error rate. sometimes one gets confused with this scheme, because
you might find application protocols which require parity bits, because
they handle them. but then these are just additional data bits to the data
link layer.

ge