Possible new feature for next Photoshop

On Thu, 20 Oct 2011 00:18:47 +0100, Bruce
wrote:

Martin Brown wrote:

XP was an unusually good vintage and there are still corporates
running it even today. Vista was almost still-born and no amount of PR
fluff and infinite budget advertising hype could resurrect it.


Was it not the case that XP wasn't all that great, but Vista was so
really, really bad that it made XP look good?

The last truly stable version of Windows was NT v4.0. Discuss.

It was at about that time that the ever increasing flood of updates
began to flow from MS. Windows 2000 originally was intended to be NT5
and can trace a direct lineage back to NT3.

Regards,

Eric Stevens

On Thu, 20 Oct 2011 00:18:47 +0100, Bruce
wrote:
[]
The last truly stable version of Windows was NT v4.0. Discuss.

It was at about that time that the ever increasing flood of updates
began to flow from MS. Windows 2000 originally was intended to be NT5
and can trace a direct lineage back to NT3.

Regards,

Eric Stevens

Windows-7 has been pretty stable here, in both 32-bit and 64-bit variants,
although it needs some routine jobs or services disabled if you have some
real-time critical hardware (I have one single core system receiving 75 GB
of satellite data per day). The memory management and cache management is
better on systems more recent than NT4, not to mention new device support
and 64-bit operation and enhanced driver security. Many of the problems
attributed to Windows come from poorly-written 3rd party drivers.

I would not wish to go back to NT4, although I do still run one Windows
2000 SP4 PC, and its only failure was the PSU one Christmas Day! Had to
pop down to Maplins on Boxing Day to get a replacement

David

tony cooper wrote:

My entry received the lowest score I've ever received in a
competition: 70. The judges said it didn't have enough blur.

It wasn't a good photo for the theme anyway. I came up with it the
last day and shot it in my garage a few hours before the deadline. I
couldn't think of a good subject. (We submit online)

http://tonycooper.smugmug.com/Other/...10-05-1-XL.jpg

I *like* the idea, since the parts are moving at varying speeds,
leading to varying blur.

But that's not great macro photography.

BugBear

On 2011-10-20 00:18:47 +0100, Bruce said:

Martin Brown wrote:

XP was an unusually good vintage and there are still corporates
running it even today. Vista was almost still-born and no amount of PR
fluff and infinite budget advertising hype could resurrect it.


Was it not the case that XP wasn't all that great, but Vista was so
really, really bad that it made XP look good?

The last truly stable version of Windows was NT v4.0. Discuss.

In my experience NT 3.51 and NT 4.0 server editions were the most
stable and could run happily on a machine with only 32 MB of RAM and a
386 CPU. Indeed, IBM made a series of server chassis for these OSs and
at least one vendor of state-of-the-art network switches used NT 4
server for control and admin. The only failures I ever experienced with
dozens of systems were hardware failures: most often hard disk drives,
secondly RAM, thirdly motherboards. Not one system failed during
endless power cycling tests thereby saving the cost of UPSs at each
site.

The 386 and later CPUs have 4 privilege rings, but most OS designers
use only ring 0 (kernel) and ring 3 (user). This may have been
acceptable for NT 4, but is totally unacceptable for the complexity of
more recent OSs including modern UNIX and Linux systems. Even OS/2 used
3 of the rings (ring 2 was used for operations such as user-mode I/O).

As a practical example, Sony BMG published CDs with copy protection and
DRM, which installed a rootkit onto the machine without the user being
aware of it (a rootkit hijacks part of the OS kernel and is extremely
difficult to detect). If the OS had been designed properly with driver
code running at ring 1 instead of 0, this would have been impossible,
as would much other malware.

The only thing that should be allowed to change kernel code is a vendor
supplied patch or update. User installed software should never be able
to modify kernel code, not even if the user is a member of the system
administrator group of users.

The next biggest downfall of NT 4 and later was having Explorer as the
shell. Explorer should run as a client of an extremely well engineered
(therefore robust) shell. E.g. some user programs install CBT hooks in
order to provide keystroke capture and injection to other applications
for remote control aka record and replay macro facilities. This forces
the injection of a DLL and application code into the workspace of
_every_ windowed user mode application running on the machine,
including the shell. An error in that one application will bring down
the whole user session. Even with error-free code, it prevents the
proper and expected serialization of tasks.

The third biggest problem with modern OSs is the user experience. Many
users get so frustrated when logged in correctly as a non-privileged
user that they use the machine logged into an admin account. There is
simply no excuse for this incredibly poor user experience in MS, UNIX,
and Linux. From NT 4 the OS has functions to create extra workstation
and desktop objects, each with their own security contexts. UNIX and
Linux have always worked on the principle: a user is a client, their
desktop is a client of the window server, the window server is a proxy
for that user and is independent of all other users. I.e. all 3 systems
cater for having more than one user logged in at the same time
therefore they all have the ability to run, say, the Web browser as
"Guest" so malware cannot harm anything other than the temporary Guest
account session - any damage caused is erased at logoff.

Until the OS designers use all 4 privilege levels provided by the CPU
and MS creates a robust shell _for_ Explorer and the user apps to all
share, we are stuck with increasingly slow, unstable and insecure
systems.

And my final observation is that while OSs continue to be written in C
and applications written in C++, we are doomed. This is insanity: at
the very least, core OS modules and all code that parses user input
should be written in Pascal. Pascal has inherent run-time range and
boundary checking and works in harmony with the x86 series (almost zero
overhead) instruction to implement these checks - buffer overflow (aka
buffer overrun) injection attacks would be impossible yet they are
still causing vulnerabilities in modern software. C was, and still is,
a jack of all trades and master of none. I read that the nuclear power
station up the coast from me has two and a half million lines of C code
controlling it - it doesn't take a rocket scientist to figure out that
there are probably quite a few bugs in that amount of C code

On 2011-10-20 12:28:44 +0100, Pete A said:

On 2011-10-20 00:18:47 +0100, Bruce said:

Martin Brown wrote:

XP was an unusually good vintage and there are still corporates
running it even today. Vista was almost still-born and no amount of PR
fluff and infinite budget advertising hype could resurrect it.


Was it not the case that XP wasn't all that great, but Vista was so
really, really bad that it made XP look good?

The last truly stable version of Windows was NT v4.0. Discuss.

In my experience NT 3.51 and NT 4.0 server editions were the most
stable and could run happily on a machine with only 32 MB of RAM and a
386 CPU. Indeed, IBM made a series of server chassis for these OSs and
at least one vendor of state-of-the-art network switches used NT 4
server for control and admin.

Oops, NT 3.51 required an i386 whereas NT 4 required an i486 for its
64-bit interlocked exchange instructions so the IBM server chassis had
an i486 not an i386. I'd forgotten those native instructions were
absent in the i386 (which required the overhead of bus lock and unlock
instructions to surround multi-word code).

"Pete A" wrote in message
news:2011102012284430046-pete3attkins@nospamntlworldcom...

In my experience NT 3.51 and NT 4.0 server editions were the most stable
and could run happily on a machine with only 32 MB of RAM and a 386 CPU.

And my final observation is that while OSs continue to be written in C and
applications written in C++, we are doomed.

I pretty much agree with everything you said.

Windows 2000 was cool then Microsoft went bananas.

The C/C++ standard would benefit from being tweaked, restricted to a subset
going forward, and proper libraries instead of that template disaster.

I know it excites some people but I really hate bloat and having to relearn
useless stuff every few years. I never thought I'd say it but more girls in
computer science might help drive standards and interoperability instead of
all this autistic crap.

--
Charles E. Hardwidge

On 2011-10-20 13:18:08 +0100, Charles E. Hardwidge said:

"Pete A" wrote in message
news:2011102012284430046-pete3attkins@nospamntlworldcom...

In my experience NT 3.51 and NT 4.0 server editions were the most stable
and could run happily on a machine with only 32 MB of RAM and a 386 CPU.

And my final observation is that while OSs continue to be written in C and
applications written in C++, we are doomed.

I pretty much agree with everything you said.

Windows 2000 was cool then Microsoft went bananas.

I was a great fan of 2000 workstation. If one went to the trouble of
disabling non required system services and performing the MS endorsed
registry tweaks to improve DNS and TCP client security, it was very
stable and quite secure OS, hence it was one of my favourite OSs. My
all-time favourite was a system of Linux servers I'd built and had them
hand-tweaked (security hardened) by a computer security guru - he also
provided a document outlining the potential security issues remaining
after the tweaks were applied. Experts such as him are few and far
between so they can charge upwards of £2000/day - out of the question
for the home user, but there is no excuse for corporations not to use
such experts.

The C/C++ standard would benefit from being tweaked, restricted to a subset
going forward, and proper libraries instead of that template disaster.

At one stage (a long time ago) _every_ template and library was riddled
with bugs. Even just the C++ file open ios flags were not implemented
properly leading to all sorts of weird application failures.

For mission critical work, I stuck with Pascal and either wrote
libraries from scratch or modified source code provided freely by
others who gave the disclaimer "... for any use whatsoever." That level
of open sharing was typical of the Pascal community. It's a pity that
attitude didn't prevail - now all we get is the likes of Apple and
Samsung trying to ban each other from selling devices, which is not
only counter to our best interests, it is plain silly/childish
behaviour from supposedly professional companies.

Remember BT claiming it invented the hyperlink therefore tried to
patent it? Luckily, common sense prevailed in this instance.

I know it excites some people but I really hate bloat and having to relearn
useless stuff every few years. I never thought I'd say it but more girls in
computer science might help drive standards and interoperability instead of
all this autistic crap.

I couldn't agree more

Many readers may not like this, but my engineering experience taught me
that the best decision makers (managers) were predominantly female.
They were prepared to consider changing their opinion/decision based on
solid empirical evidence whereas males frequently took the stance "I
hear what you're saying, but 'we' have decided to do it this way so
that is the way you will do it." When I became self-employed, my reply
to that was always "Then find another contractor because I refuse to
earn money from supplying crap." Luckily, I never lost a day's work by
being totally professional. Also luckily, I retired when I did because
that work ethic has become as extinct as the dinosaurs.

On Thu, 20 Oct 2011 12:28:44 +0100, Pete A
wrote:

On 2011-10-20 00:18:47 +0100, Bruce said:

Martin Brown wrote:

XP was an unusually good vintage and there are still corporates
running it even today. Vista was almost still-born and no amount of PR
fluff and infinite budget advertising hype could resurrect it.


Was it not the case that XP wasn't all that great, but Vista was so
really, really bad that it made XP look good?

The last truly stable version of Windows was NT v4.0. Discuss.

In my experience NT 3.51 and NT 4.0 server editions were the most
stable and could run happily on a machine with only 32 MB of RAM and a
386 CPU. Indeed, IBM made a series of server chassis for these OSs and
at least one vendor of state-of-the-art network switches used NT 4
server for control and admin. The only failures I ever experienced with
dozens of systems were hardware failures: most often hard disk drives,
secondly RAM, thirdly motherboards. Not one system failed during
endless power cycling tests thereby saving the cost of UPSs at each
site.

The 386 and later CPUs have 4 privilege rings, but most OS designers
use only ring 0 (kernel) and ring 3 (user). This may have been
acceptable for NT 4, but is totally unacceptable for the complexity of
more recent OSs including modern UNIX and Linux systems. Even OS/2 used
3 of the rings (ring 2 was used for operations such as user-mode I/O).

As a practical example, Sony BMG published CDs with copy protection and
DRM, which installed a rootkit onto the machine without the user being
aware of it (a rootkit hijacks part of the OS kernel and is extremely
difficult to detect). If the OS had been designed properly with driver
code running at ring 1 instead of 0, this would have been impossible,
as would much other malware.

The only thing that should be allowed to change kernel code is a vendor
supplied patch or update. User installed software should never be able
to modify kernel code, not even if the user is a member of the system
administrator group of users.

The next biggest downfall of NT 4 and later was having Explorer as the
shell. Explorer should run as a client of an extremely well engineered
(therefore robust) shell. E.g. some user programs install CBT hooks in
order to provide keystroke capture and injection to other applications
for remote control aka record and replay macro facilities. This forces
the injection of a DLL and application code into the workspace of
_every_ windowed user mode application running on the machine,
including the shell. An error in that one application will bring down
the whole user session. Even with error-free code, it prevents the
proper and expected serialization of tasks.

The third biggest problem with modern OSs is the user experience. Many
users get so frustrated when logged in correctly as a non-privileged
user that they use the machine logged into an admin account. There is
simply no excuse for this incredibly poor user experience in MS, UNIX,
and Linux. From NT 4 the OS has functions to create extra workstation
and desktop objects, each with their own security contexts. UNIX and
Linux have always worked on the principle: a user is a client, their
desktop is a client of the window server, the window server is a proxy
for that user and is independent of all other users. I.e. all 3 systems
cater for having more than one user logged in at the same time
therefore they all have the ability to run, say, the Web browser as
"Guest" so malware cannot harm anything other than the temporary Guest
account session - any damage caused is erased at logoff.

Until the OS designers use all 4 privilege levels provided by the CPU
and MS creates a robust shell _for_ Explorer and the user apps to all
share, we are stuck with increasingly slow, unstable and insecure
systems.

And my final observation is that while OSs continue to be written in C
and applications written in C++, we are doomed. This is insanity: at
the very least, core OS modules and all code that parses user input
should be written in Pascal. Pascal has inherent run-time range and
boundary checking and works in harmony with the x86 series (almost zero
overhead) instruction to implement these checks - buffer overflow (aka
buffer overrun) injection attacks would be impossible yet they are
still causing vulnerabilities in modern software. C was, and still is,
a jack of all trades and master of none. I read that the nuclear power
station up the coast from me has two and a half million lines of C code
controlling it - it doesn't take a rocket scientist to figure out that
there are probably quite a few bugs in that amount of C code

C has been described as a 'read only' language.

Even the person who wrote it can later find it impossible to decypher
how it works.

Regards,

Eric Stevens

Pete A wrote:
On 2011-10-14 21:17:04 +0100, Charles E. Hardwidge said:

[...]
Microsoft made a huge deal out of dropping 16 bit support in 64 bit OS, and
after reading through the (junior) development team's report explaining why
spotted where they'd made some mistakes. The reality is they just weren't
capable enough to figure out how to do it and the clock was ticking.

Money, money, money.

Exactly. Intel CPUs maintain 16-bit emulation mode available for any
_competent_ OS designer to support. The fact that most hardware vendors
no longer provide a floppy disk drive does not mean the CPU is
incapable of running 16-bit DOS programs. To me, this shows an
incredible feat of backwards compatibility engineering by Intel and the
incredible level of incompetence of some OS vendors.

You should realize that that backwards compatibility comes at a price.
Higher prices and lower performance.

--
Ray Fischer | Mendocracy (n.) government by lying
| The new GOP ideal

On 2011-10-23 03:04:19 +0100, Ray Fischer said:

Pete A wrote:
On 2011-10-14 21:17:04 +0100, Charles E. Hardwidge said:

[...]
Microsoft made a huge deal out of dropping 16 bit support in 64 bit OS, and
after reading through the (junior) development team's report explaining why
spotted where they'd made some mistakes. The reality is they just weren't
capable enough to figure out how to do it and the clock was ticking.

Money, money, money.

Exactly. Intel CPUs maintain 16-bit emulation mode available for any
_competent_ OS designer to support. The fact that most hardware vendors
no longer provide a floppy disk drive does not mean the CPU is
incapable of running 16-bit DOS programs. To me, this shows an
incredible feat of backwards compatibility engineering by Intel and the
incredible level of incompetence of some OS vendors.

You should realize that that backwards compatibility comes at a price.
Higher prices and lower performance.

Windows has both.