At 11:36 AM +1000 8/8/01, Bernard Robertson-Dunn wrote:
>
>Would it not be a good idea for an operating system to separate code and
>data? Like in some of the operating systems I programmed in the mid 1970s?
Broadly speaking most OS's still do ... into the heap and the stack.
Windows problem is that where something loads and how it activates is
pretty well dependent on its 3 letter extension ... a pretty
primitive way to determine what is code and what is data ... and the
fact that ActiveX controls enable code available from within data to
be activated by various engines/code in the heap (VBScript, Macro
engines etc) in a pretty well uncontrolled manner and with full
privileges. There's no no-execute policy in place ... in fact quite
the contrary . ActiveX controls specify an almost unlimited execute
policy between different applications. 'It's a feature, not a bug'.
The facts that there is no process control or data channel monitoring
on a networked Wintel machine are also a worry. Broadly speaking the
internal data channels will accept most any input from outside as
long as it arrives on the right port/protocol. There's also no or
very little file integrity checking in the OS, and OS files and
libraries can be over-written with gay abandon.
>
>Windows is still a toy operating system and it sometimes worries me that
>the world is becoming so dependent on such a poorly architected
>environment, and that includes both the hardware and software. The 1970s
>systems had separate registers for the operating system and user space. A
>benefit for both security and performance.
It does the job. It has relatively few overheads. It's cheap. It has
an OK GUI for the average Joe Public. For games and the like it has
probably one of the best and most comprehensive API's around now
(DirectX). As single user machines, Wintel machines are fine. Hey
I've got one that I use for client testing server apps (and games).
But no way will I use it to connect to the Net.
The problems come when you use them as servers, or when you network
them ... and all those 'cross application features' in Wintel
applications and the OS suddenly raise a number of security concerns.
You have to remember that until 5 or so years back, MS was primarily
a NETBUI networked company (Hey, it was their own protocol, much like
AppleTalk/Appleshare is Apple's). TCP/IP and WAN is a whole new ball
game to them ... and I really don't think they've come to grips with
pervasive networking and WAN and the security issues that raises.
Thier share of the server market probably reflects this. They
basically give away their core server products (and charge like
wounded bulls for the peripheral products that you need to actually
make them do something useful) .... but they still don't even
approach Apache, MySQL, PHP and the like for an installed base. (And
Apache is an order of magnitude harder to set up than MSIIS for
example.)
I think that's part of the problem ... the MS front ends are
deceptively easy to setup, but the back end is where all the problems
arise and requires a hell of a lot of care to first get right, and
second to regularly patch as new security vulnerabilities arise.
Total cost of ownership stats for MSIIS and Apache would be
interesting ... I haven't seen any firm figures, but MSIIS seems to
require far more ongoing maintenance and servicing than a typical
Apache installation (and is a hell of a lot less stable).
To my mind if MS wants to continue with its current architecture and
its limitations it should do one thing. License and provide a copy of
Nortons Antivirus and Nortons Firewall as part of each and every
default Windows install ... together with paid up subscriptions for
two or three years updates and scripted automatic updates from
Symantec (or whoever) for every networked MS machine. Then MS could
say it was doing something to alleviate the problem ... it wouldn't
stop it, but it would alleviate it. As it is, each incident is
costing $100's of millions or billions (I personally lost ... and had
to replace ... a motherboard last year to the Wintel CIH virus -
thanks mainly to an omission on my part and the brain dead policy of
Intel allowing unrestricted writes to the BIOS of all motherboards
produced over the last 36 months "Hey it's another feature.") and a
heap of inconvenience and aggravation for millions of innocent users.
Just my 2 cents worth ...
Regards,
-- ********************************** Some days you're a bug, other days a windshield. **********************************
This archive was generated by hypermail 2.1.1 : Fri Aug 31 2001 - 03:10:03 EST