Re: Split the kernel sources

Wakko Warner (wakko@animx.eu.org)
Sun, 14 Nov 1999 13:34:23 -0500


> what do you think about splitting the kernel sources ?
>
> i'm fed up with these (>10MB) big-fat kernel sources. i will never
> need most of this stuff. so i thought about splitting the kernel
> source tree in many small parts:

> * base system (main makefiles, include files,...) -
> for kernel and for modules compiling without the whole kernel source.
>
> * main kernel
> * architecture dependent stuff
> * device drivers (splitted in many small packages ...)

> i could imagine, we have an base core which is needed for compiling the
> kernel
> or only for modules. (this could be splitted in packages for several
> platforms)
>
> people who only want to compile some drivers need only this package and
> the driver package.

How about the people who (like myself) would much rather download that 10mb
file once, and use patches the rest of the time? Actually, it sounds like
the kernel source being split like this would mean that they'd have to have
that many more patch files for each kernel.

I had the idea about splitting the kernel until I realized what I said
above. I have mostly x86 machines here, and one sparc. Once compressed,
the difference is trivial. I just did a compare (tar gzipped) of all the
dirs in arch. 2.4mb Kill all but i386, you'll save 2.2mb. about 11
minutes at 28.8k.

Drivers: 9mb compressed. I guess most of the stuff is here (I'm using
2.3.27 to base this on) Want my list on that one? File systems, another
meg. ANything I left out, 4mb compressed.

How would the build scripts work? would they have to know what is there or
have a directory that it sources in all files to know what to do?

Don't misunderstand what I said here. It would be a nice idea, but for the
maintainers, it might be more of a hassle to work with. If the patches were
files that patched each of these things, One like myself would have to
download all relevent patches (instead of blindly getting patch-x.x.x.gz).

-- 
 Lab tests show that use of micro$oft causes cancer in lab animals

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/