Path: utzoo!attcan!uunet!mitel!sce!cognos!dgbt!gandalf!carr
From: carr@gandalf.UUCP (Dave Carr)
Newsgroups: comp.dcom.modems
Subject: Re: Request for information (again)
Summary: Why manufacturers claim 4:1
Message-ID: <2617@gandalf.UUCP>
Date: 5 Oct 89 14:16:18 GMT
References: <998.251E4CDB@zswamp.fidonet.org>
Lines: 23

In article <998.251E4CDB@zswamp.fidonet.org>, Geoffrey.Welsh@p0.f171.n221.z1.fidonet.org (Geoffrey Welsh) writes:
> 
>    I'm not sure it will be that high. I know that it's possible (the latest 
> version of PKZIP will compress a text file to 15-20% of its original size), 
> but compression "on the fly" is less powerful, especially if you don't want 
> to incurr big delays by keeping enough data in memory to make powerful 
> analyses on them before sending them off.
>  
True.  But most modem manufactures include the *compression* gained by
stripping the stop and start bits (and sometimes parity) to get figures as
high as 4:1.  Starting at 10-bit characters, you only have to get down to
2.5 bits/character to claim 4:1.  This is quite possible.  PKZIP is working
on 8-bit data.

The BT-LZW also learns quicker, and adapts even after the tables fill.
It also has a CLEAR TEXT mode, so uncompressable data doesn't expand.
I don't know the inner workings of PKZIP, so I can't compare this to
V.42 bis.

The delay question can be addressed by upping the speed at the terminal
side of the modem to 38.4 or higher.  The modem will usually buffer data
to get higher compression.  That is, if the compression is low, the buffers
will fill to a higher level, and therefore compression will improve.