🚀 go-pugleaf

Path: kodak!grodberg
From: grodberg@kodak.UUCP (jeremy grodberg)
Newsgroups: mod.telecom
Subject: Re: error correcting modems
Summary: Error correction with notification is superfluous.
Message-ID: <733@kodak.UUCP>
Date: 9 Mar 87 23:48:12 GMT
References: <8702230043.AA02476@alliant.Alliant.COM>
Reply-To: grodberg@kodak.UUCP (jeremy grodberg)
Distribution: world
Organization: Eastman Kodak Co, Rochester, NY
Lines: 44

In article XXX steckel@alliant.Alliant.COM (Geoff Steckel) writes:
>
>A number of vendors have introduced "error correcting" modems recently.
>I have heard of none that notify the user that correction has taken place.
>...  From experience and reading in networking, this is a disaster waiting to
>happen.


   From my experience with error correcting modems, notification of error
correction would be pointless.  First, at high speeds (4800 baud and up) over
normal phone lines, there is error correction taking place very frequently
(more often then every 1000 characters).  Thus the notification would be so
frequent as to be routinely ignored, and therefore pointless, totally leaving
out the fact that such notification would necessarily make the modem non-
transparent, which would greatly reduce it usefullness.
   Besides, notification that error correction has take place has nothing to do
with passing along incorrect data.  Incorrect data comes from so many errors
ocurring in the same block that the error correcting scheme gets fooled
into thinking some erroneous data was correct because "it checks."  True,
some good data may be trashed as well, but it is also possible that the 
mistakes will fall so that no error correction takes place at all.
Certainly, the fact the error correction took place is not at all a good
indicator that bad data was passed, and in many error correction schemes, it
is not possible to know when an uncorrectable error has occurred (although
there are some that can catch a high percentage of such errors, none can guard
against all possible error sequences).

   It may be adviseable to use a low-power error correction scheme or error
detection scheme with ECC modems, but I'm not sure.  We have been routinely
transmitting around 750K of binaries a day at 4800 baud (with error correction 
and data compression) via Motorola S-records, which provide very good error
detection.  We have been doing this 5 days a week for over two months, and have
NEVER detected an uncorrected tranmission error, even though we frequently have
error correction taking place in every block (512 characters).

   If an error correcting scheme can promise me no more than one uncorrected
error every six months of use, I'll be happy.  Nothing is perfect, and it is
dangerous to make things almost perfect, because it leads to humans putting
too much faith in such systems (It has to be right, the computer says so.).

          -Jeremy Grodberg

Usenet: ...rochester!kodak!grodberg
Arpa: 	grodberg@kodak or kodak!grodberg@rochester