Re: slink is gone, goals for potato?
On Wed, Mar 10, 1999 at 07:12:07PM -0500, Ben Pfaff wrote:
> Stephen Crowley <crow@debian.org> writes:
>
> Why not for packages? I think the majority of people have faster machines
> rather than slower. How long are we going to keep this up? I mean in 5 years
> from now are we still gonna be worried about the 4 people out there running
> 386s? I hope not. No offense, but there is no need for this. A 486 is gonna
> be slow whether its gzipped or bzip2ed, what about us poor modem users? Not
> everyone has T1 to their house you know.
>
> Before you start saying things like that, try considering some facts.
> I downloaded the entire slink base section from a local mirror,
> extracted the tarballs, then recompressed them with gzip -9v and bzip2
> -9v. The results:
>
> blp:~/debian-cd/bin(0)$ du base.tgz
> 13897 base.tgz
> blp:~/debian-cd/bin(0)$ du base.bz2
> 13374 base.bz2
> blp:~/debian-cd/bin(0)$
>
> i.e., bzip2 is approximately 4% better. Do you want to slow down
> everyone's installs for a 4% improvement in download speed?
>
> Okay, now let's do the same thing with the source archives. Results
> this time:
>
> blp:~/debian-cd/src(0)$ du src.tgz/
> 27089 src.tgz
> blp:~/debian-cd/src(0)$ du src.bz2/
> 21729 src.bz2
> blp:~/debian-cd/src(0)$
>
> for a 25% bzip2 advantage. So it *might* be worth it for source code,
> but definitely not for binaries.
Please recognize that the decompression performance of bzip2 is
substantially worse. For the source archive of the 2.2.1 kernel, the
decompression goes from 4.4 seconds to 29.2 seconds--300MHz PII, four
samples averaged.
IMHO, even the 25% space improvement may not be worth the order of
magnitude performance loss.
- O
Reply to: