Re: [RFD] New driver system (was Adding new syscalls via modules)

Alexander_Maryanchick%STORACTIVE@storactive.com
Tue, 6 Jul 1999 13:52:33 +0400


>> 1. ext2compr, Undelete Tools, KGI,... are patches.
> - I've never heard of a kernel patch for undelete-tools.
http://amadeus.upr.clu.edu/~undelete/welcome.html
> - ext2compr is a patch to ext2fs which is itself a module
Unfortunately, not only to ext2fs.
> - KGI... can't it be made a module as well ?
No. There is KGIcon, but it is different.

> CD-ROM cache amounts to CacheFS which can be made as a module.
Cool.
> HDD exists just fine.
Do you mean 2.2 or 2.3? :-)
Some kernel versions are broken because of one new feature!
The kernel must not depend of cache.
> web-acceleration ??
??? :-)

> I'm not trying to say that more extendibility wouldn't be good, but that
> there already is a good amount of flexibility and that what is missing
> is not only/necessarily missing in the core itself.
Agree, but the main problem is not features - TIME !
Here for example a triller about 'FS forks' (or another cool) feature:
1. Months (years?) of discussion (now) about is it relly neded.
This time is lost for projects using ugly replacements for the
'forks'.
2. Implementation (I mean it as necessary because *people* want it).
2.5.14 'forks' patch (10 kb)
2.5.15 - 2.5.35 kernels are broken because of amount of unneeded
dependencies.
ext3fs bugs - till 2.6.8 kernel.
months are lost !
3. Testing (Mindcraft-2001)
2.7.3 'forks' patch (1 kb)
2.7.4 - 2.7.11 kernels are broken...
Although the above is a science fiction, it based on reality :-(.

Another large problem is _distribution_.
Neither home nor corporate users like to recompile the kernel.
Therefore commercial driver developers will follow NT too.

It is much more than just some features.

Best regards

Alexander.

-
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/