Simplicity has a future

[Richard Heathfield]

Richard Bos said:

Oh ja? Wel, _jouw_ mammie draagt legerkistjes!

Prosze mowic po angielsku. Dziekuje!

 

[Richard Tobin]

The chance of a memory error resulting in the need to reinstall is
so low it is not usually worth the money. I don't think I've seen
a case attributable to it in 15 years, and none for certain before
that. If we'd put ECC memory in all our machines it would have
added up to a huge amount over the years.

[/QUOTE]

You don't know, because the non-ECC machine simply ignores the
error.

I do know, because I haven't had any cases of computers needing a
reinstallation where the cause wasn't known to be something other
than a memory error.

-- Richard

 

[CBFalconer]

I do know, because I haven't had any cases of computers needing a
reinstallation where the cause wasn't known to be something other
than a memory error.

You don't know, because apart from out and out hardware failure the
only reason to reinstall is fouled files. These can be caused at
any time during copying by a dropped bit in a buffer. Disk
compaction is especially vulnerable. Remember, a memory error will
not be immediately noticed, but may be delayed by weeks, months,
even years.

--
"Our enemies are innovative and resourceful, and so are we.
They never stop thinking about new ways to harm our country
and our people, and neither do we." -- G. W. Bush.
"The people can always be brought to the bidding of the
leaders. All you have to do is tell them they are being
attacked and denounce the pacifists for lack of patriotism
and exposing the country to danger. It works the same way
in any country." --Hermann Goering.

 

[Richard Tobin]

I do know, because I haven't had any cases of computers needing a
reinstallation where the cause wasn't known to be something other
than a memory error.

[/QUOTE]

You don't know, because apart from out and out hardware failure the
only reason to reinstall is fouled files. These can be caused at
any time during copying by a dropped bit in a buffer.

I do known, because the only times I have had to reinstall have been
when a disk failed. I understand that people using Microsoft Windows
are often advised to reinstall their systems, but for the rest of us
it is a very rare occurrence.

-- Richard

 

[Andrew Poelstra]

You don't know, because apart from out and out hardware failure the
only reason to reinstall is fouled files. These can be caused at
any time during copying by a dropped bit in a buffer. Disk
compaction is especially vulnerable. Remember, a memory error will
not be immediately noticed, but may be delayed by weeks, months,
even years.

What exactly does POST do on non-ECC memory? Would cheap memory be the
reason that it runs so fast?

(This doesn't fit into the conversation nicely, but neither does the
conversation fit into the group nicely).

 

[Walter Roberson]

I do known, because the only times I have had to reinstall have been
when a disk failed. I understand that people using Microsoft Windows
are often advised to reinstall their systems, but for the rest of us
it is a very rare occurrence.

And what caused the disks to fail? Are you 100% sure that those
failures were not caused by software problems, nor contributed to
by software problems?

For example, if one of the disks failed due to stiction, then is
it certain that the automatic preventative maintenance cycles against
stiction never had difficulties due to corrupted memory bits?

 

[Richard Tobin]

For example, if one of the disks failed due to stiction, then is
it certain that the automatic preventative maintenance cycles against
stiction never had difficulties due to corrupted memory bits?

We're discussing the costs and benefits of fitting ECC memory in your
PC, not what memory is used on disk drive controllers.

-- Richard

 

[Richard Tobin]

What exactly does POST do on non-ECC memory? Would cheap memory be the
reason that it runs so fast?

It writes and reads back various bit patterns.

-- Richard

 

[Walter Roberson]

We're discussing the costs and benefits of fitting ECC memory in your
PC, not what memory is used on disk drive controllers.

First off, we aren't discussing costs and benefits of fitting ECC
memory into a "PC": we are discussing cost and benefits of fitting
memory error detection/correction schemes (of whatever ilk) into
computers in general.

Secondly, preventative maintenance of drive problems is not
necessarily controlled by the drive controllers. That's not
a theoretical "not necessarily". My SGI IRIX boxes have
a cron job that wakes up twice a week (2 A.M. Sunday and Thursday)
and sends the commands to spin down the drives and 2 minutes later
sends the commands to spin them back up again. If the SGI boxes
did not happen to have ECC, then a memory bit error could corrupt
the low-level controller commands (with potentially harmful results),
or a memory bit error could simply abort cron (so that the commands
never run), or a memory bit error could corrupt the filesystem
DRAM cache of the anti-stiction program (again with potentially
harmful results.)

Of course then there's always the possibility that a regular
disk write command got corrupted into something less benign
(e.g., format sector using the data block as a template) that
led to a disk failure.

If your memory isn't being error-checked then it is often pretty hard
to prove that memory problems were -not- the root cause of a failure.

 

[Richard Tobin]

First off, we aren't discussing costs and benefits of fitting ECC
memory into a "PC"

CBFalconer asserted that the cost of reinstalling an operating system
made it worth-while spending more for ECC memory; I responded that it
would not have been worth it for me; he asserted that I could not know
that; I responded that I knew the cause of the very small number of
reinstallations that I have done or known about in the last 15 years.
I made no claim about the value of ECC memory in other circumstances.

Secondly, preventative maintenance of drive problems is not
necessarily controlled by the drive controllers. That's not
a theoretical "not necessarily". My SGI IRIX boxes have
a cron job that wakes up twice a week (2 A.M. Sunday and Thursday)
and sends the commands to spin down the drives and 2 minutes later
sends the commands to spin them back up again.

None of my computers have done that. In fact, the computers in
question were not left on all the time so their disks were frequently
turned off.

Of course then there's always the possibility that a regular
disk write command got corrupted into something less benign
(e.g., format sector using the data block as a template) that
led to a disk failure.

My disk failures were not of that kind either.

No-one disputes that there is an advantage to ECC memory in many
situations, but the claim that it is always false economy not to
use it is not borne out by my experience.

-- Richard