Re: Are we losing users to Gentoo?
On Thu, Nov 21, 2002 at 01:05:32PM +0100, Adrian 'Dagurashibanipal' von Bidder wrote:
> The somewhat dubious performance aspect notwithstanding, as was
> pointed out, the archives would grow HUGE. How many architectures
> would Debian have? i[3456]86, Athlon, perhaps a few odd ones like
> Cyrix M2 and the older or newer K6. MIPS is at least 5 CPUs or so,
> 68k, SPARC and ARM, too. Don't know anything about the other arches,
> but I suspect it's similar. So Debian should grow to more than 50
> supported cpu variants?
it's bad enough that we have all the kernel-image packages precompiled
for every i386 variant, with and without SMP.
i remember a year or so ago when i complained about this worthless practice i
said that it would end up consuming hundreds of megabytes - i was told that was
ridiculous, it would never happen.
at the moment, the various i386 kernel-image (plus kernel-headers and
kernel-pcmcia-modules packages) in pool/main/k/kernel-image-* are taking up
about 310MB.
the images for 2.4.19 alone use up 96MB:
falls:/home/ftp/pub/mirrors/debian/pool/main/k# l kernel-image-2.4.19-i386
total 96724
drwxr-xr-x 2 root root 4096 Nov 20 19:25 ./
drwxr-xr-x 208 root staff 8192 Nov 20 19:26 ../
-rw-r--r-- 1 root root 4020582 Nov 19 09:17 kernel-headers-2.4.19-386_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4021768 Nov 19 09:17 kernel-headers-2.4.19-586tsc_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4023562 Nov 19 09:17 kernel-headers-2.4.19-686-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4022292 Nov 19 09:17 kernel-headers-2.4.19-686_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4021870 Nov 19 09:17 kernel-headers-2.4.19-k6_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4023436 Nov 19 09:17 kernel-headers-2.4.19-k7-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 4022496 Nov 19 09:17 kernel-headers-2.4.19-k7_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3928442 Nov 19 09:17 kernel-headers-2.4.19_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3956 Nov 19 09:17 kernel-image-2.4-386_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3974 Nov 19 09:17 kernel-image-2.4-586tsc_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3982 Nov 19 09:17 kernel-image-2.4-686-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3974 Nov 19 09:17 kernel-image-2.4-686_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3974 Nov 19 09:17 kernel-image-2.4-k6_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3984 Nov 19 09:17 kernel-image-2.4-k7-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 3972 Nov 19 09:17 kernel-image-2.4-k7_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9248886 Nov 19 09:17 kernel-image-2.4.19-386_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9188280 Nov 19 09:17 kernel-image-2.4.19-586tsc_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9467880 Nov 19 09:17 kernel-image-2.4.19-686-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9190712 Nov 19 09:17 kernel-image-2.4.19-686_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9098818 Nov 19 09:17 kernel-image-2.4.19-k6_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9639914 Nov 19 09:17 kernel-image-2.4.19-k7-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 9363962 Nov 19 09:17 kernel-image-2.4.19-k7_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 236048 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-386_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 235364 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-586tsc_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 238818 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-686-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 234786 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-686_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 234362 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-k6_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 242078 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-k7-smp_2.4.19-4_i386.deb
-rw-r--r-- 1 root root 237812 Nov 19 09:17 kernel-pcmcia-modules-2.4.19-k7_2.4.19-4_i386.deb
since each kernel-image usually gets up to at least the -4 or -5
release, each one consumes 4 or 5 times as much bandwidth on each
mirror.
and this is to save a handful of users who probably wouldn't even notice
the difference between an i386 kernel and an i686 or athlon optimised
kernel from the hassle of compiling their own kernel on their own
machine. compiling a kernel really isn't that hard to do if you use
make-kpkg.
IMO, those who have the faster machines, who actually care about the few
percentage points of performance can and should compile their own kernels. for
everyone else, an i386-optimised kernel is good enough because that will boot
and run on every i386 variant.
craig
--
craig sanders <cas@taz.net.au>
Fabricati Diem, PVNC.
-- motto of the Ankh-Morpork City Watch
Reply to: