[plug] RE: Linux backward compatability baggage

[David Campbell]

To:             	plug at linux.org.au
Subject:        	Re: [plug] Microsoft ask users to crack win2000 site (fwd) 
Date sent:      	Sat, 07 Aug 1999 07:36:38 +0800
From:           	John Summerfield <summer at os2.ami.com.au>
Send reply to:  	plug at linux.org.au

> > Mike Holland wrote:
> > 
> > > Amen. A consequence of feature-bloat, backwards compatibility, ...
> > 
> > Just out of interest, what baggage is linux carrying? It has backward
> > compatibility to the i386 architecture obviously. Is this problem
> > entirely solved by being able to select the code you want to go into the
> > kernel, when you configure?
>
> Assuming all the system time's kernel, to reduce its overhead would have 
> little overall benefit; for what users actually do, optimising X would be 
> better. In my case, there might be some advantage in optimising perl - it 
> consumes most of the user and nice values above: that's where real work's 
> done.
> 
> In this context, I refer to gnerating CPU-specific code with compiler 
> options. Better algorithms will always beat compiler optimisation, as 
> would upgrading the CPU: that box has a P133: the M/b can take up to 
> P233MMX and equivalent clones.

Providing a clear distinction between kernel and program is kept 
systems will be:
   a) stable
   b) fast

Linus has kept the division very clear as demonstrated by his 
willingness to change the kernel structure (for the sake of system 
stability) and ignoring the up-roar by the people using the Andrews 
File System (AFS) module.

Linus still has to comply with the standards which define the 
interface between applications and kernel (as defined by the POSIX 
group) except where there is a very good reason not to (eg: an 
optional specification is just plain stupid).

Microsoft in successfully attempting to push MS-Windows as "the 
operatating system" has allowed all and sundry to write "kernel 
extensions" (otherwise known as "Virtual Device Drivers" or VXDs).

As a result we have these "low cost" Taiwan digital device companies 
(scaners, printers, modems - you know the "Windows only" type of 
equipment). These device drivers have never been developed in a 
"bunker mentality" development environment where interitent software 
failure is a crime equal to murder. Unfortunately it is true of any 
operating system that the kernel side is only as strong as the 
weakest link (eg: the worst device driver).

The problem of backward compatibility is the fact that back doors 
exist in all operating systems. Microsoft appears loathed to shut 
these back doors hence creating defacto standards rather than 
properly addressing the problems which created them in the first 
place.

David Campbell
=======================================================
campbell at torque.net
"This is not an office, rather Hell with fluorescent lighting"