NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/sys/sun/hardware/3946

Internet Message Format | 1992-08-19 | 9.3 KB

Xref: sparky comp.sys.sun.hardware:3946 comp.dcom.modems:12274
Newsgroups: comp.sys.sun.hardware,comp.dcom.modems
Path: sparky!uunet!bif!mike
From: [e-mail] (Mike Heins)
Subject: SUMMARY: Does anybody REALLY need 115.2K baud?
Message-ID: <[e-mail]>
Organization: Central Data Corp., Champaign, IL
Date: Wed, 19 Aug 1992 19:36:04 GMT
Lines: 211

Thanks to everyone who responded (25+ replies by mail or post).  Almost
everyone thought that 115.2K baud or FASTER is needed now or will be in
the near future.  The most common application and equipment mentioned was
UUCP or SLIP file transfer on a Telebit WorldBlazer.  V.FAST and Codex
modems were mentioned as well.

At least one user thought we shouldn't be worrying our heads:

 ----------------------------------------------------
| From James W. Adams   --   NIH Scientific Computing Resource Center
|
| This thread begs the question of using an antiquated interface designed
| for half-duplex modems at <= 300 BPS to run at throughput exceeding that
| of the original ARPANET backbone.
|
| Serial has too much overhead for this to be efficient and flow control
| is still a problem.  Several posters have raised the issue of
| availability.  Nearly everyone who has a Sun has both a SCSI and an
| Ethernet connection.  Ethernet and SCSI cards for PCs are also widely
| available.  Macs have SCSI and Ethernet is easily available.
|
| It's time to retire the UART and move on.
 ----------------------------------------------------

The only problem with this is many of us still have to connect to
phone lines -- I for one would LOVE a direct Internet connection,
but...

A couple of people were unaware that speeds greater than 19200 were
supported on the internal Sun serial ports.

 ----------------------------------------------------
| From [e-mail] Sat Aug 15 09:19:24 1992
|
| I know 19200 is definitely a throttle on my Worldblazer; I got over
| 3000 CPS with Turbo PEP on uncompressed ASCII (postscript, as it
| happens).
 ----------------------------------------------------

 ----------------------------------------------------
| From: Gerard Hynes <[e-mail]>
|
| How about the CODEX 3266 V.fast modems I have here?
| DTE speeds up to 115k and link speeds of 24k with V.42bis data
| compression on top.....(Yes, they are *very* fast!)
|
| BTW, how can I get 115k out of the serial port on a SS2?
| Any help would be appreciated......
|
| Currently, all I can get is 19.2k....but I only got the modems
| yesterday and am testing/configuring them on 2 PC's (ugh!) which
| I can push out to 115k.
|
| PPP between 2 SS2's with these modems and sufficently high DTE
| speeds should be usable for almost any application. (X...maybe!)
|
 ----------------------------------------------------

38.4K baud is supported on ttya and ttyb as it is, as long as you don't
rely on the configuration done by the boot PROM.  The standard baud
code table in <ttydef.h> supports up to 38.4K baud.

The main problem with running at these speeds on the non-intelligent
internal serial ports is the several *thousand* interrupts the CPU
must take per second.  The resulting overhead does not leave much room
for much other useful work, and application throughput suffers greatly.

Additional comments are collected below.

All in all, it seems that a product offering real 115.2K baud support
(and throughput) will be appreciated by the Sun market.  Stay tuned!

Mike Heins
------------------------
Central Data Corporation    "Makers of the scsiTerminal Server"
1602 Newton Drive        [phone]
Champaign, IL 61821        [e-mail]   or  ..!uunet!cendata!mike

 ----------------------------------------------------
| From: bang![e-mail] (Bret A. Marquis)
|
| >From the Telebit Worldblazer manual:
|
| "Worldblazer supports DTE interface speeds of 115,200 bps.  This allows
| data transmission rates of more than 70,000 bps with data compression
| and PEP".
|
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Steve Simmons)
|
| Supposedly the Telebit Worldblazer runs approx 24K modem-to-modem.
| With 3:1, that 72K.  V.FAST is coming, who knows what actual thruput
| will be.
|
| You can use other things besides async modems on those ports.
|
| And two years from now, 115.2 may not be fast enough.
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Bill Mayhew)
|
| Without going into a mathematical discussion, Shannon's theorem
| implies that the maximum information rate that one should expect
| from an average voice grade telephone circuit is around 24,000 bits
| per second.  The Telebit Worldblazer modem promises around 23,000
| bits per second, half duplex.  These are uncompressed base rate
| figures.  With factal algebra or something new, who knows what
| we'll be able to achieve.  Even LZW compression can obtain up to
| 3:1 ratios on plain text, so you need more than 57.6Kbuad (a
| standard interface rate) to get full use from the modem.  115Kbuad
| is the next standard rate that is commonly used.
|
| Driving the modem async is not terribly efficient at such high data
| rates.  Of course, the Worldblazer also supports SDLC sync
| connections.  Of course, also, sync connections are pretty alien to
| the Unix world.
|
| If one does the compression in the host, an interface good enough
| to support the base rate will suffice.  Many modern modems,
| however, offload the compression to a procesor in the modem
| itself.
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Jeff Wasilko)
|
| The only modem I know of that runs at 115kbaud is the Telebit Worldblazer.
| So, to drive the modem at it's best, you'd need to get at least 90kbaud
| worth of data to it. 115kbaud was the next highest standard, so Telebit
| picked it.... You might want to ask the same question in comp.dcom.modems,
| as it was recently discussed over there....
|
| I think that since modems (and data comression rates) are getting
| faster, we'll all need faster serial ports...
|
 ----------------------------------------------------

 ----------------------------------------------------
| From a user whose name is withheld by request:
|
| The major reason I would want that fast of a rate is to run SLIP and
| have near-ethernet response times for file transfers, etc..
|
| I have talked to several people who run X-windows from home into work.
| They say that response times are pretty sluggish, and it's just now
| getting usable with the current high-end consumer modems.
|
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Joel Berendzen)
|
| The reason I might use a 115Kbaud modem, if there were one, is to run X-window
| clients on an X-terminal at home.  The reasons this looks attractive are (1)
| the suddenly low cost of X terminals (~$1.5K), (2) My unwillingness to go back
| to character-mode terminals after adapting to windowing systems, (3) the
| high bandwidth required for running X protocols.
|
|
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Andrew Burt)
|
| Pure speculation on my part here, but (a) modem speeds will increase
| such that 115kb will eventually be obsolete, so "yes" I would like a product
| now that does it; and (b) eventually (I hope) serial will be replaced with
| something better (faster, more standardized, just as cheap/easy/available,
| etc.), also obsoleting 115kb boards, but I don't see that in the near term,
| so again "yes" I'd like such a board now.  Also given that just getting 38kb
| in a serial board for a unix box seems less than a given, I'd like to see
| the pressure applied so that high speed connections are made easier to
| obtain.
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Andreas Schulz)
|
| The Standard V.42bis tells you about a typical 4:1 data compression and
| some new asynch modems go faster than 19.2K baud. Turbo PEP is 23.0K baud
| and the new V.fast proprosal sometimes talk about 28.0K baud. And
| 28.0 x 4 is definitely mor than 57.6K, maybe 112Kb :-). For Turbo PEP
| you need 23.0 x 4 = 92.0Kb , for the USR HST 16.8Kb and Zyxel 16.8Kb you
| need 16.8 x 4 = 67.2Kb , all more than 57.6K baud. And the only next
| standardised baud rate is 115.2.
|
 ----------------------------------------------------

 ----------------------------------------------------
| From: [e-mail] (Greg Kemnitz)
|
| If you are trying to do any sort of video stuff through the phone lines, you
| need speeds like this, preferably faster.  Also, the faster your modem runs,
| the better PPP runs.  Certainly if you are just doing TTY logins to your host
| from work, you may not need such speed, but people will do much more
| interesting things with modems in the very near future once these high speeds
| become commonplace.
|
| The recent FCC rule change allowing phone companies to compete with the cable
| companies could also open up new markets for high-speed modems.
|
 ----------------------------------------------------