>Optimizing bunzip is not worth the time/effort IMHO. Because it's a
>CPU resource hog and only provides a couple of percentage points
>reduction in the compressed file size compared to gzip. By far a
>better solution is to ship .gz install files.
+1 I've measured it and it's somewhere between 5-8%.
But gunzip runs rings around it when it comes to performance
>The original decision to ship .bz2 install files was flawed IMHO.
It was based on media size constraints; still an issue on SPARC nearing
the single layer DVD limit. But it was, I think, because of CD sizes.
There are two things I'd like to measure:
- gunzip vs bunzip2 installation times
- not writing the contents file during install
(I have a prototype which does this safely but it needs to
be added into the pkg* tools)
Together, I think, this will speedup installs by a factor 2-3 (depending
on the speed of CPU and disks) and upgrades even more)
Another thing that I'd like to see investigated is allowing the ramdisk to
be paged out during install when the installer has configured the disks
and added swap. This will give the installer and OS quite a bit more
cache space to further improve performance, specifically on memory
starved systems.
Paging the ramdisk is not much different from running from a disk; it
should certainly not affect speed all that much.
Casper
