Re: dpkg binary dbase (was Re: Debian vs. Red Hat)
On 7 Sep 2000, John Hasler wrote:
> Bruce Sass writes:
> > The result is still human readable and editable with any text editor, if
> > you know the codes. The "special dpkg editor" would just make life
> > easier for those not wanting to look up or learn any codes.
>
> Ok, but I'm not sure that it would be significantly faster then a
> well-designed text database.
Neither am I. A little experiment using pppconfig
(dpkg -s pppconfig > pppconfig.orig.txt, then cutting out the
description) yielded a file size of 392 bytes, encoding it...
pppconfig:2.0.5:111:90:999:2:5
D ppp (>= 2.3.7), whiptail | dialog
C /etc/init.d/dns-clean 0e085f68dd877596f115bdbaaecf5050
C /etc/ppp/ip-up.d/0dns-up 267ad65788adb5fec899a5ed77bffaf4
C /etc/ppp/ip-down.d/0dns-down 88c6ee9d50c2ac5b4de19af26300d956
yields 248 bytes - a 36.7% size decrease. How typical that is of the
rest of the packages, and wether the decrease in disk I/O offsets the
now required cases and table lookups -- ?
The first part of the above question could be handled with statistical
methods, a large enough random sampling of the Debian archive, and a
program to encode entries in the existing DB. The second part would be
a complicated equation that depends on coding style and the operation
being performed, very messy to analyse.
later,
Bruce
Reply to: