Path: utzoo!utgpu!watmath!clyde!att!pacbell!ames!pasteur!agate!saturn!ssyx.ucsc.edu!ulmo
From: ulmo@ssyx.ucsc.edu (Brad Allen)
Newsgroups: comp.mail.uucp
Subject: Re: Making /usr/lib/uucp/paths a _binary_ database
Keywords: pathalias UUCP smail
Message-ID: <5916@saturn.ucsc.edu>
Date: 5 Jan 89 05:46:42 GMT
References: <486@fallst.UUCP> <490@gonzo.UUCP>
Sender: usenet@saturn.ucsc.edu
Reply-To: ulmo@ssyx.ucsc.edu (Brad Allen)
Lines: 23

>Other schemes are possible, but most seem to me to bne performance
>killers too.

RANDOM MUMBLING MODE:

   19643   39286  753869 paths.txt
# of !'s:  51132
		  261207 paths.txt.Z (compact and pack did 40%)

2 bytes for a unique # (less using Hoffman coding techniques)
6 bytes for sitename (if you compress this a little ... comparing is
		      the same cost after 1 quick compression)
8 bytes for its hops

16 bytes total per entry ... oh, that's 320000 bytes.

Amazing.  Not much of an advantage.  That's not even counting the search into
the filename.  Well, take some of this back.  If on every sitename you gave
diff from last sitename, this figure might drop to 14 bytes?  10 bytes with
path compression?  Then the indexes would make it a lot.  So, as a rough
estimate, pressing really hard, we could get it down to under 200000 bytes.

Time to replace UUCP with SLIP ...