lkml.org 
[lkml]   [1999]   [Feb]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: proper place to discuss kernel 'bloatedness'?
In <19990206235524.08094@fireplug.net> Stuart Lynne (sl@fireplug.net) wrote:
>> | > Couldn't the kernel be split up into a server kernel and workstation
>> | > kernel?
>> | And where would you split the kernel? What makes a linux machine a server is
>> | not the kernel, but the various daemons SERVING some kind of service. The
>> | kernel "merely" provides a medium for those daemons - networking, file
>> | systems, access to hardware etc.
>> +--->8
>>
>> Most of the difference between "workstation" and "server" is parameter tuning,
>> not kernel code. This is true even of NT, where (aside from bundled software)
>> the primary difference is registry entries which tune various parameters.
>>
>> And most of the "kernel bloat" is device drivers. I can *almost* see some
>> point in splitting off e.g. SCSI or sound trees --- except that every time I
>> do something like that with my own code, I end up regretting it. It's one
>> of those "nice in theory" things....

> I would suggest splitting the kernel source into that which might be
> required to boot and that which is not required to boot and is easily
> handled as a module.

How about devices required to boot but easily handled as a module ? Like module
for handling my AIC-7890 ? "Required to boot" and "easily handled as a module"
are not opposited !
NPU emulation -- required to boot and not easily handled as module
AIC-7890 -- required to boot but easily handled as module
procfs -- not required to boot but not easily handled as module
sb16 -- not required to boot and easily handled as module
All four variants are there in kernel just now :-)

> I.e. almost all block device drivers stay with the main kernel tar ball.

What for ?

> Almost all character devices and all network drivers could easily get moved
> into a Linux Modules source tar ball that could have an independant (to the
> kernel) life.

"Easily" ? Are you joking ? With moves like planned move of rename() from
filesystem-level to VFS-level are forced to redesign all modules (AFS peoples
are heppy to hear about this move, I know :-))) ! So if module will be
saparated from main kernel this will become mantainance nightmare (like PCMCIA
just now).

> Is there any compelling reason for being able to compile ethernet drivers
> into the main kernel? They work fine as modules. Ditto for serial, isdn,
> sound, etc.

Some peoples just hate modules and it's just easily to mantain consistency
with current big tarball then with proposed scheme...





-
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/

\
 
 \ /
  Last update: 2005-03-22 13:50    [W:0.239 / U:0.012 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site