[comp.dcom.modems] HST's vs. Telebits

jessea@dynasys.UUCP (Jesse W. Asher) (04/11/90)

In article <1438@mindlink.UUCP> a186@mindlink.UUCP (Harvey Taylor) writes:
>In <134@dynasys.UUCP>, jessea@dynasys.UUCP (Jesse W. Asher) writes:
>} The number of _unix_ sites running HSTs vs. running Telebits are
>} negligible.
> Jesse,
>     Can you document this? I tried to get this kind of info from our
> sysop, but he says he doesn't have the uucp maps, the only source I
> know from which to extract such data. The folklore I have heard is that
> HST has the BBSes, PEP has Unix & Hayes has business, but I have never
> seen figures for Unix sites (or business). I pulled some data for Fido
> sites from a local board (see below).


Unfortunately, I'm only going on experience.  I orignally came from the dos
world so I feel I'm qualified to talk about both sides.  HST's seem to be
used mainly by people from or familiar with the dos world while Telebits
seem to be used mainly by those from or familiar with the unix side.  That's
not to say that unix sites don't run HSTs and dos sites don't run telebits.
But, because of the problem the HSTs had with the uucp g protocol for so long,
telebits have become the standard for the unix world.

This is neither scientific or anything else along those lines, but I did the
following:  I grepped through the uucp maps and found -

	4 lines referencing "HST" in regard to the modem (there is an HST time
		zone).
	5 lines referencing "USR" in regard to the modem.
		- two of these sites also had Telebits.
	66 lines referencing "[tT]elebit".

That's a pretty big margin - even with a large margin for error.  If you
doubt the results, anyone is welcome to produce more accurate ones.  But I
don't think its going to make much difference in the final outcome.  If you
take the high number of 5 out of 66, you get 7.6%.  Not a very large number
and definitely not a standard.

I don't think V.32 will be the great mediator either.  I think a protocol
that has the robustness and speed of the Telebits and HSTs will have to
become the standard for such a standard to exist.  It's very difficult to
accept the performance of V.32 once one has used a telebit (I'm sure the
same applies to HSTs as well).


-- 
Jesse W. Asher - Dynasys - (901)382-1705     Evening: (901)382-1609 
6196-1 Macon Rd., Suite 200, Memphis, TN 38134  
UUCP: {fedeva,chromc,autoz}!dynasys!jessea

karl@ddsw1.MCS.COM (Karl Denninger) (04/11/90)

In article <139@dynasys.UUCP> jessea@dynasys.UUCP (Jesse W. Asher) writes:
>
>This is neither scientific or anything else along those lines, but I did the
>following:  I grepped through the uucp maps and found -
>
>	4 lines referencing "HST" in regard to the modem (there is an HST time
>		zone).
>	5 lines referencing "USR" in regard to the modem.
>		- two of these sites also had Telebits.
>	66 lines referencing "[tT]elebit".

Note that we could probably account for at least 10 of those "Telebit"
entries ourselves with our Usenet and email neighbors.  We are hardly 
connected to 1/6th of the Telebit-using Unix sites!

I suspect that most of the sites who have them don't publicize it.  Why
bother, unless you want to have unsolicited data calls?

--
Karl Denninger (karl@ddsw1.MCS.COM, <well-connected>!ddsw1!karl)
Public Access Data Line: [+1 708 566-8911], Voice: [+1 708 566-8910]
Macro Computer Solutions, Inc.   "Quality Solutions at a Fair Price"

jimb@faatcrl.UUCP (Jim Burwell) (04/13/90)

jessea@dynasys.UUCP (Jesse W. Asher) writes:

>In article <1438@mindlink.UUCP> a186@mindlink.UUCP (Harvey Taylor) writes:
>>In <134@dynasys.UUCP>, jessea@dynasys.UUCP (Jesse W. Asher) writes:
>>} The number of _unix_ sites running HSTs vs. running Telebits are
>>} negligible.
>> Jesse,
>>     Can you document this? I tried to get this kind of info from our
>> sysop, but he says he doesn't have the uucp maps, the only source I
>> know from which to extract such data. The folklore I have heard is that
>> HST has the BBSes, PEP has Unix & Hayes has business, but I have never
>> seen figures for Unix sites (or business). I pulled some data for Fido
>> sites from a local board (see below).

>This is neither scientific or anything else along those lines, but I did the
>following:  I grepped through the uucp maps and found -

>	4 lines referencing "HST" in regard to the modem (there is an HST time
>		zone).
>	5 lines referencing "USR" in regard to the modem.
>		- two of these sites also had Telebits.
>	66 lines referencing "[tT]elebit".

Put us (faatcrl) in for 12 of those Telebits.  We have 12 Telebit T2500s,
which gives us V.32 (9600 baud/full duplex), and PEP (19.2Kbaud/adaptive
duplex).  

Believe me.  Telebit OWNS the Unix market.  This is because they were the only
people to think of marketing a high speed modem which gets good results with
the UUCP 'g' protocol.  To get maximum file transfer throughput from a PEP 
connection when using UUCP 'g', Kermit, Xmodem, and Ymodem, all of which rely
on a full duplex connection, Telebit put "File Transfer Protocol Support" into
it's modems.  You enable this by setting an S register (S111 on a T2500) to
the value for the protocol you'll be using for that call.  The modem detects
when the file transfer starts and "spoofs" the file transfer between the 
computer and the modem, while the two Telebits simply talk PEP between 
themselves (PEP is error correcting, of course).  This allows PEP to allocate 
nearly all if it's channels to outgoing data for the sender, and incoming data 
for the receiver (of course).  Thus, the maxmum throughput of PEP can be 
realized even when using a full duplex file protocol.  If the Telebits didn't 
spoof, the modem would allocate some channels for input, and some for output,
since there is a lot of handshaking going on.  Ymodem wouldn't be THAT bad,
since it only ACKs every 1K block sent.  But 'g', and 'Kermit' (kermit for
the most part) send VERY short (32 to 64 byte) blocks, acks would be coming
back from the receiver very often, and them modem would probably divide the
send/receive channels evenly, thus dropping from almost 19.2KBaud throughput
to 9600 baud throughput!  And since UUCP is locked into using the 'g' protocol,
Telebits are the only answer.  (Too bad it doesn't use a streaming protocol
like Zmodem or something)

I have heard that USR now makes an HST Dual-Standard-U or something for Unix
systems, which also has 'g' spoofing (not sure about Kermit or [XY]modem). 
But even then, I probably wouldn't buy a HST, since they're not nearly as
good on noisy/bad phone lines as the Telebits under PEP (PEPs a trooper on
noisy lines!).  And we have some noisy lines here.  :-)  Owell.  I'll stop now.

C'ya,
Jim
-- 
James S. Burwell
UUCP:  ...!rutgers!faatcrl!jimb        Internet:  jimb@faatcrl.UUCP
"Idiots... " - Anonymous

larry@nstar.UUCP (Larry Snyder) (04/13/90)

In article <1319@faatcrl.UUCP>, jimb@faatcrl.UUCP (Jim Burwell) writes:
> I have heard that USR now makes an HST Dual-Standard-U or something for Unix
> systems, which also has 'g' spoofing (not sure about Kermit or [XY]modem). 
> But even then, I probably wouldn't buy a HST, since they're not nearly as
> good on noisy/bad phone lines as the Telebits under PEP (PEPs a trooper on

The US Robotics solution is to use software running on the OS - completly
replacing the uucp protocol with their own version of it to push the HST
up to 1650 cps.  The problem is that with their software running, normal 
UUCP connections can not be establishe

-- 
...!iuvax!ndmath!nstar!larry  -or-  larry@nstar

michael@stb.info.com (Michael Gersten) (04/15/90)

In article <1319@faatcrl.UUCP> jimb@faatcrl.UUCP (Jim Burwell) writes:
>on a full duplex connection, Telebit put "File Transfer Protocol Support" into
>it's modems.  You enable this by setting an S register (S111 on a T2500) to
>the value for the protocol you'll be using for that call.  The modem detects
>when the file transfer starts and "spoofs" the file transfer between the 

Exactly. This is the problem--you have to specify in a register which
protocol you are using on that call. And both sides have to specify the
same one. **ARRGGHH**.

Effectivly, it is UUCP only spoofing, unless you let your users change
their end and your end to kermit (in which case they can change your
end into a real mess).

Much better would have been either two registers (so kermit or xmodem
could be spoofed as well), or simply a toggle register (and allowing
any of the protocols supported to run), or even a bitmask (it supports
less than 8 protocols, right?). <SIGH>.

As mail says, ubluit.

		Michael
-- 
		Michael
michael@stb.info.com denwa!stb!michael anes.ucla.edu!stb!michael 
Disclaimer: It works for me.