telebit modems and unix (+ standards/compression miscellany)

[James A. Woods]

#  "first there is a mountain.  then there is no mountain.  then there is."
       -- d. leitch, after rene daumal

     telebit and unix?

for decent performance you want:

   either,
	19200 serial input hardware with modem/flow control and large silos
	-- older pdp11 serial boards and vax dz-11s need not apply.
	many workstations are weak here, e.g. the masscomp 5000 is known to
	have aggregate 19.2 serial i/o for all ports (dedicated ones
	run only at 9600), and the sun 2 manual certainly doesn't
	mention 19.2 dialup.
   or,
	some sort of rs232-to-ethernet bridge box.
   and,
	a kernel to support something better than interrupt-per-character
	input on serial lines, preferably something like the bsd 4.3
	auto-siloing method.
   plus,
	kermit/uucp software with large buffers (telebit recommends 10000
	bytes) to minimize the effects of line turnaround.
	honey danber uucp permits easy buffer size changes (e-protocol),
	but assumes an error-free channel which may not be present 
	under the conditions lauren weinstein has warned about.
	older uucp software may have problems with dialer code
	not being table-driven; the telebit command set is different
	from the hayes in subtle ways.  again, ask lauren about this.
	i know of the existence of a one-line telebit description for
	honey danber uucp:

        telebit	 =W-,	"" ATZ OK\r-+++\c-OK\r ATZ "" \EATDT\T\r\c CONNECT

	buying hdb might payoff for someone instead of rewriting/recompiling
	ad-hoc dialer code.

now, the only reason i can sound properly authoritative (in the traditional
usenet style) about this without actually having run this wonder at
14kbps, is that, stuck with a pdp 11/70 for modem connections, my
workgroup at ames can't run at full max with the device.  the evaluation
unit we played with did run reliably at lower speeds for the couple
of months we had one.  i can only be frustrated that a lowly ibm pc
can be setup quickly for fast file xfer much easier than supposedly more
capable cpu/software combinations.

     telebit and standards?

the 9600 baud standard is weak -- fallback to drastically lower speeds are what
dooms other modems in any fair race over typical non-leased dialup lines.
the technology of the telebit is far superior (data rate over phone lines
is actually 18bps (non-compressed) including error correction bits)
i wish them well in a standards battle, though they may have to give away
the technology to gain committee votes.

     telebit and compression?

some manufacturers (see electronics, 4/28/86) offer repeat character compression
to boost rates, but this pales with lempel-ziv.  telebit indeed knows about
the unix compress utility; future m68k integration is not farfetched given that
the telebit roms are infused with 60000 lines of C code.  the only concern
i have about this would be to disable compression on any modem buffer
which won't compress -- since some folks already squeeze data via software.

the marketing advantage of claiming 30 kps over dialup lines would be
compelling.

ames!jaw