Xref: utzoo comp.dcom.modems:5309 biz.comp.telebit:37
Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!apple!vsi1!teraida!ditka!kls
From: kls@ditka.UUCP (Karl Swartz)
Newsgroups: comp.dcom.modems,biz.comp.telebit
Subject: Re: poor uucp performance - help! (LONG)
Message-ID: <21377@ditka.UUCP>
Date: 23 Feb 90 19:01:27 GMT
References: <125@cdss.UUCP> <1990Feb17.063412.18455@rancho.uucp> <959@codonics.COM>
Reply-To: kls@ditka.UUCP (Karl Swartz)
Organization: Inaction Central, San Jose, California
Lines: 39

In article <959@codonics.COM> bret@codonics.COM (Bret Orsburn) writes:
>In article <1990Feb17.063412.18455@rancho.uucp> rock@rancho.uucp (Rock Kent) writes:
>>4.  Make sure that you have compression turned off.  You don't want to
>>    be compressing already compressed news batches.

>OK, I'll bite: Why not?

>Surely, compression will be less effective the second time, but is there
>any good reason to disable it? Do you have any data to demonstrate that
>throughput is *decreased* by enabling compression for a compressed newsfeed?

Compressing random binary data will often make a file larger since
there are no patterns for the compression algorithm to eliminate,
and assuming the algorithm does a good job (which it does) a file
that has already been compressed will almost certainly grow since
they now have the overhead of the compression tables in addition
to the uncompressible data.

Compress is smart enough to recognize that it's efforts were wasted
and refrains from replacing the original with a larger compressed
file, but this only works if it has control over input and output
files.  Obviously this doesn't apply for a pipeline, which is in
essence what's being done in a Telebit, all it can do is emit the
larger stream of bytes.

Since you wanted hard data, I made up a batch consisting of the
usual "#! cunbatch" header and a compressed (16 bit) copy of the
first Northern California file from comp.mail.maps (u.usa.ca.2).
I then compressed the file using both 12 and 16 bit compression
(I believe the Telebits use 12 bit compression) with results as
follows:

    file		bytes	compression  time @800CPS
    original		35755	    n.a.	  44.7
    compress -b12	49206	  -37.61%	1:01.5
    compress -b16	47053	  -31.59%	  58.8

Doesn't look like I'll be running off to enable compression on
my TrailBlazer soon.