All,
Interesting. I found a different story. I took a disk utilization of
the /usr/src/linux directory, first looking at the arch subdirectory:
21 ./i386/boot/compressed
8 ./i386/boot/tools
96 ./i386/boot
16 ./i386/mm
369 ./i386/math-emu
10 ./i386/lib
194 ./i386/kernel
696 ./i386
9 ./alpha/boot/tools
22 ./alpha/boot
25 ./alpha/lib
10 ./alpha/mm
204 ./alpha/kernel
48 ./alpha/math-emu
323 ./alpha
13 ./sparc/boot
369 ./sparc/kernel
81 ./sparc/lib
139 ./sparc/mm
40 ./sparc/prom
652 ./sparc
3 ./mips/boot
12 ./mips/mm
190 ./mips/kernel
5 ./mips/lib
224 ./mips
38 ./ppc/mm
54 ./ppc/kernel/include/elf
55 ./ppc/kernel/include
244 ./ppc/kernel
58 ./ppc/boot/compressed
69 ./ppc/boot
9 ./ppc/lib
370 ./ppc
263 ./m68k/amiga
195 ./m68k/atari
33 ./m68k/boot/amiga
34 ./m68k/boot/atari
69 ./m68k/boot
533 ./m68k/ifpsp060
208 ./m68k/kernel
15 ./m68k/lib
44 ./m68k/mm
502 ./m68k/fpsp040
314 ./m68k/console
3 ./m68k/tools/amiga
4 ./m68k/tools
2163 ./m68k
4430 .
After removing all non-i386 items, I had:
21 ./i386/boot/compressed
8 ./i386/boot/tools
96 ./i386/boot
16 ./i386/mm
369 ./i386/math-emu
10 ./i386/lib
194 ./i386/kernel
696 ./i386
699 .
In the /usr/src/linux/include directory I found the following:
28 ./linux/modules
1129 ./linux
4 ./asm-generic
197 ./asm-i386
265 ./asm-mips
201 ./asm-alpha
255 ./asm-m68k
344 ./asm-sparc
108 ./net
137 ./asm-ppc
11 ./scsi
2653 .
After removal of non-i386 items, I had:
28 ./linux/modules
1129 ./linux
4 ./asm-generic
197 ./asm-i386
108 ./net
11 ./scsi
1452 .
When I did a "make clean," and retarred the linux-2.0 directory, I had
the following file:
-rw-r--r-- 1 root root 4771581 Jun 16 17:10 /home/ftp/pub/linux/linux-2.0-i386.tar.gz
Compared to the original source:
-rw-r--r-- 1 root root 5843677 Jun 9 11:48 /home/ftp/pub/linux/linux-2.0.tar.gz
There must be other architecture-specific files in the normal kernel
distribution; e.g., in Documentation? This is only a rather surface
attempt to size the benefits of a formal architecture-oriented
development tree. (Incidentally, I was able to rebuild the kernel and am currently
running it.)
I think a reduction of over 20% and close to 1.1 MB is a major
difference and certainly worth "the trouble," especially when we think
of the future. What happens when we add more ports. What about porting
to IBM Mainframes!
It seems to me that what is at issue here boils down to the
convenience of the kernel developers. If a Linux Guy wants to download
the whole distribution, please do. Why should everyone have to?
I think that an argument to the effect that the kernel group needs to
have all the different drivers, etc., makes sense. I just don't see
why everyone else must.
I won't keep posting. I hope people will seriously consider the
possibilities. I believe there must be a good solution. Look at the
X11R6 Distributions.
Best regards,
Paul Matthews
McLean, VA
e-mail: paul@matthews.com