Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!unisoft!hoptoad!xanth!kent
From: kent@xanth.UUCP (Kent Paul Dolan)
Newsgroups: comp.sources.bugs,news.misc
Subject: Re: sending source code
Message-ID: <3249@xanth.UUCP>
Date: Fri, 6-Nov-87 23:51:01 EST
Article-I.D.: xanth.3249
Posted: Fri Nov  6 23:51:01 1987
Date-Received: Mon, 9-Nov-87 06:20:53 EST
References: <631@louie.udel.EDU> <332@uvicctr.UUCP> <2566@umn-cs.UUCP> <467@srs.UUCP>
Reply-To: kent@xanth.UUCP (Kent Paul Dolan)
Organization: Old Dominion University, Norfolk Va.
Lines: 63
Keywords: compress directly to printable ASCII
Summary: Does anyone want me to do this?
Xref: mnetor comp.sources.bugs:426 news.misc:1108

In article <467@srs.UUCP> dan@srs.UUCP (Dan Kegel) writes:
>Roughly paraphrased:
>> It is desirable to use a compressing archiver (like zoo or arc)
>> to package groups of files for transmission.
>> However, the resulting archives must be uuencoded (or btoa'd) before
>> transmission on Usenet.
>> Why not just make the compressing archiver output in a format
>> suitable for direct transmission on Usenet?
>
>Sounds like a good idea to me.  I've always disliked having to go
>thru three steps (paste, uudecode, unarchive) to decode these postings;
>it's work I'd just as soon have a program do for me.
>
>I think that more thought needs to be given to transmitting large
>binary files over Usenet.  People often distribute documentation and
>source in this format, so the old objection "But executables don't belong 
>on Usenet" no longer applies.
>
>- Dan Kegel


A couple of months ago, I took the algorithm from the June '87 CACM
article "Arithmetic Coding for Data Compression", and recoded it under
contract into FORTRAN 77.

Due to stupidities on both sides (mostly mine, I'm afraid), I didn't
get paid, the software didn't get used, and all copies were destroyed!
However, I ended up owner of the neat algorithm additions I invented.

I had made it work, efficiently, and with one nice wrinkle.  The
original CACM algorithm encoded output bits into 8 bit bytes for
compressed data storage.  I needed to send the resulting file across a
smart communications line; i.e., the transmitted data had to be
printable.  The kermit escape encoding looked too expensive, so I made
a little switch.  It turns out that 95*95 (printable ASCII characters)
is just a bit more than 2^13, so by encoding 6.5 bits of data per byte
by doing the obvious mod, multiply and shift for each byte, I was able
to compress into printable ASCII.  The result was about a wash for
executables; they were about as big transmitted (compressed into
printable ASCII) as unencoded and uncompressed, which is probably at
least as good as the present situation.  Various kinds of text behaved
quite a bit better.

So, would it be worthwhile for me to rewrite this stuff in C, or would
someone else like to go ahead and do this, given these hints?  I, too,
think that a one step process to do this would be better.  I can do
the compression and decompression routines, and put them on the net,
if someone who does Unix systems stuff could go on from there and make
them into a uuencode-/uudecode-like utility, which is probably beyond
my skills.  I assume that the CACM algorithm is publicly usable.

Comments?

Kent, the (totally weird) man from xanth.

Running for president on a pound of caffeine, an ounce of sense, and a
program of increased exploration and exploitation of space.  Support
your (probably non-existent - get busy!) local branch of the Birthright
Party:  "The birthright of mankind is the stars!"

Hey, it's better than dwelling on your stock portfolio; at least here
you've got a chance for a laugh or two. ;-)  Yum!  Eat them plastic
chickens, brethren!  Call me when I'm elected; 'til then, I'm going to
take a nap.