Path: bga.com!news.sprintlink.net!qns1.qns.com!constellation!convex!
convex!cs.utexas.edu!swrinde!ihnp4.ucsd.edu!newshub.sdsu.edu!
mentor.sdsu.edu!mgutierr
From: mguti...@mentor.sdsu.edu (Mario Gutierrez)
Newsgroups: comp.os.linux.admin
Subject: Why is Linux monolithic?
Date: 24 May 1994 07:40:22 GMT
Organization: San Diego State University, College of Sciences
Lines: 13
Message-ID: <2rsb16$q19@pandora.sdsu.edu>
NNTP-Posting-Host: mentor.sdsu.edu
X-Newsreader: TIN [version 1.2 PL2]

Having just completed my Operating Systems course, I'm wondering why
Linux is a monolithic operating system?  Moreover, why not loadable 
device drivers?  Don't get me wrong I like Linux very much, I'm just
curious?  I think the device drivers being built into the system is a
big drawback, especially when you have a driver which patches against
a certain version of the Linux kernel.  And if you want to add another
driver which patches against a different version of the Linux kernel
don't you run into problems?  Can anyone shed some light on this?


--
mario l gutierrez
mguti...@mentor.sdsu.edu

Newsgroups: comp.os.linux.admin
Path: bga.com!news.sprintlink.net!get.hooked.net!decwrl!hookup!
news.moneng.mei.com!howland.reston.ans.net!EU.net!uknet!cf-cm!cybaswan!
iiitac
From: iii...@uk.ac.swan.pyr (Alan Cox)
Subject: Re: Why is Linux monolithic?
Message-ID: <1994May24.140451.14420@uk.ac.swan.pyr>
Organization: Swansea University College
References: <2rsb16$q19@pandora.sdsu.edu>
Date: Tue, 24 May 1994 14:04:51 GMT
Lines: 14

In article <2rsb16$...@pandora.sdsu.edu> mguti...@mentor.sdsu.edu 
(Mario Gutierrez) writes:
>Having just completed my Operating Systems course, I'm wondering why
>Linux is a monolithic operating system?  Moreover, why not loadable 
>device drivers?  Don't get me wrong I like Linux very much, I'm just
>curious?  I think the device drivers being built into the system is a
>big drawback, especially when you have a driver which patches against

Linux isn't what I would describe as monolithic in structure. It just happens
to be linked as one object file. There is also work afoot on making large
parts of the kernel into loadable modules to avoid the relink phase.

Alan

Newsgroups: comp.os.linux.admin
Path: bga.com!news.sprintlink.net!hookup!usc!howland.reston.ans.net!
pipex!sunic!EU.net!news.eunet.fi!news.spb.su!gsp!news-service
From: j...@usoft.spb.su (John Lynn Roseman)
Subject: Re: Why is Linux monolithic?
X-Return-Path: gsp!usoft!jlr
Organization: UrbanSoft AO (St-Petersburg)
Date: Wed, 25 May 1994 10:37:00 GMT
Message-ID: < m0akklZ-01831xD@usoft>
Sender: news-serv...@arcom.rcom.spb.su
Lines: 9

Alan Cox (iii...@uk.ac.swan.pyr) writes in <1994May24.140451.14...@uk.ac.swan.pyr>:
> Linux isn't what I would describe as monolithic in structure. It just happens
> to be linked as one object file. There is also work afoot on making large
> parts of the kernel into loadable modules to avoid the relink phase.

Linux is definitely monolithic in the context of a system like the Hurd
(as are virtually all OS's in use today).
Loadable modules are a convenient short term fix, but the migration of 
Linux onto Mach may be a more promising long term strategy.

Newsgroups: comp.os.linux.admin
Path: bga.com!news.sprintlink.net!demon!uknet!cf-cm!cybaswan!iiitac
From: iii...@uk.ac.swan.pyr (Alan Cox)
Subject: Re: Why is Linux monolithic?
Message-ID: <1994May25.160536.15779@uk.ac.swan.pyr>
Organization: Swansea University College
References: < m0akklZ-01831xD@usoft>
Date: Wed, 25 May 1994 16:05:36 GMT
Lines: 24

In article < m0akklZ-01831xD@usoft> j...@usoft.spb.su (John Lynn Roseman) 
writes:
>Linux is definitely monolithic in the context of a system like the Hurd
>(as are virtually all OS's in use today).
>Loadable modules are a convenient short term fix, but the migration of 
>Linux onto Mach may be a more promising long term strategy.

Be careful what you class as monolithic. Its traditionally used to refer
to systems where any bit of the kernel goes off to call any other bit of
the kernel. Linux isn't like that - nor is it like the hurd idea where
you can just lob stuff in at random. 
Mach in itself doesn't help at all - at that level all mach does is provide
a standardised interface method between the different modules - you could
easily write this bit from scratch for Linux - Mach also doesn't really help
you in the changing interfaces issue that modules suffers from - you send
crap down a port the other end won't follow it.
When it comes to running multiple real systems, multiple CPU's and a vague
(all too vague in mach3) amount of portability then MACH comes into its own -
but to harness it needs a system built to use the ability - maybe that'll
be HURD. (Round here everyone keeps calling Hurd Linux/NT - 8) )

Alan

Newsgroups: comp.os.linux.admin
Path: bga.com!news.sprintlink.net!hookup!swrinde!gatech!
howland.reston.ans.net!pipex!sunic!news.funet.fi!news.eunet.fi!
news.spb.su!gsp!news-service
From: j...@usoft.spb.su (John Lynn Roseman)
Subject: Re: Why is Linux monolithic?
X-Return-Path: gsp!usoft!jlr
Organization: UrbanSoft AO (St-Petersburg)
Date: Thu, 26 May 1994 07:47:00 GMT
Message-ID: < m0akklZ-211xiaD@usoft>
Sender: news-serv...@arcom.rcom.spb.su
Lines: 16

Alan Cox (iii...@uk.ac.swan.pyr) writes in <1994May25.160536.15...@uk.ac.swan.pyr>: 
> Be careful what you class as monolithic. Its traditionally used to refer
> to systems where any bit of the kernel goes off to call any other bit of
> the kernel. 
I always thought monolithic referred to a single executable in which 
everything is linked into one module.  But this is just semantics.

> Mach in itself doesn't help at all - at that level all mach does is provide
> a standardised interface method between the different modules - you could
> easily write this bit from scratch for Linux  . . .
The FSF, OSF, DEC and IBM seem to be convinced that it does help.
What's the point of rewriting it from scratch?

>          (Round here everyone keeps calling Hurd Linux/NT - 8) )
The Hurd SHOULD be Linux/NT, but how can it happen unless Linux migrates
onto Mach?

Newsgroups: comp.os.linux.admin
Path: bga.com!news.sprintlink.net!hookup!swrinde!pipex!uknet!cf-cm!cybaswan!iiitac
From: iii...@uk.ac.swan.pyr (Alan Cox)
Subject: Re: Why is Linux monolithic?
Message-ID: <1994May26.124116.11142@uk.ac.swan.pyr>
Organization: Swansea University College
References: < m0akklZ-211xiaD@usoft>
Date: Thu, 26 May 1994 12:41:16 GMT
Lines: 27

In article < m0akklZ-211xiaD@usoft> j...@usoft.spb.su (John Lynn Roseman) writes:
>> Mach in itself doesn't help at all - at that level all mach does is provide
>> a standardised interface method between the different modules - you could
>> easily write this bit from scratch for Linux  . . .
>The FSF, OSF, DEC and IBM seem to be convinced that it does help.
>What's the point of rewriting it from scratch?

In all these cases they are looking at using Mach not just as a way of splitting
up a program but a way of sharing multiple subsystems properly across user
services. Windows NT, IBM'sPowerPC microkernel and HURD are all about running
lots of things together as the OS. With the exception of hurd however its 
very much a drive to get multiple monolithic OS on one machine at once. Merely
being a modular kernel can be done much more efficiently than using MACH - 
running tons of simultanous subsystems and sharing multiple CPU's between OS
layers is something where MACH becomes appropriate - its a question of scale.

>
>>          (Round here everyone keeps calling Hurd Linux/NT - 8) )
>The Hurd SHOULD be Linux/NT, but how can it happen unless Linux migrates
>onto Mach?

A Linux syscall emulator wouldn't be diabolically hard, and a ext2fs for Mach
seems quite viable looking at all MIB's documentation. 

Alan