Re: OFFTOPIC: Re: bzip2 for kernel dists?

Mitch Davis (mjd#NOSPAM@nsmd.aus.hp.com)
Fri, 12 Dec 1997 14:41:09 +1100

Jordan Mendelson wrote:
>
> So, bzip2 yeilds better compression, it's free (no patent restrictions on
> algs, etc), and it's available in binary & source form for Linux & a number
> of other OSs. What's keeping kernels from being distributed like this?

Other people's mileage may vary, but I have found bunzip2 to be
absolutely hideously SLLOOOWWW. The following test shows bunzip2
to be 10 times slower to decompress than gunzip:

mjd@taco [~/tx] time cat gimp-0.99.15-data-extras.tar.* > /dev/null

real 2.3
user 0.0
sys 0.2
mjd@taco [~/tx] time cat gimp-0.99.15-data-extras.tar.* > /dev/null

real 0.1
user 0.0
sys 0.1
mjd@taco [~/tx] time bunzip2 -t gimp-0.99.15-data-extras.tar.bz2

real 26.8
user 26.2
sys 0.1
mjd@taco [~/tx] time gunzip -t gimp-0.99.15-data-extras.tar.gz

real 2.1
user 1.9
sys 0.0

Note this is under HP-UX, but the results are very similar under Linux.
I guess it depends on whether you have CPU cycles to burn.

Mitch.

-- 
| mailto:mjd#NOSPAM@nsmd.aus.hp.com  | Not an official view of:  |
| mailto:mjd#NOSPAM@alphalink.com.au | OpenView Telecom Division |
| Remove the #NOSPAM to send me mail | Hewlett Packard Australia |