Opinions on safety of dropbox ???

[Mike O'Dell]

If you don't care whether the data is right,
why are you bothering to store it?

YES - the experimental evidence from Sun and Google and
many others is that PC hardware fucks up REALLY often.
What REALLY PISSES ME OFF is that Intel is so damn duplicitous
about it. They say personal desktops don't need ECC, but
then look at their "engineering workstations" and they implore
people to use *only* ECC machines.

so is getting your taxes right a big deal or not?
Intel doesn't want to get sued because a bridge fell down
when they got the wrong answer because of no ECC on memory.
I guess they don't expect to get sued when your bank balance
is screwed-up? Or the dosage on your medication is wrong?

The REAL reason is that with a small number of exceptions at
the very high end, the ONLY differences between their "server"
processors which all have ECC and "desktop" processors which
almost none have ECC is, well, whether they have ECC and
that the server processor costs a material integer multiple
of the price of the "desktop" processor.

the ECC story is about maintaining an indefesible price difference.

oh wait - there *are* i5 and i7 processors WITH ECC, but amazingly,
they ONLY come in ball-grid-array packages which must be reflowed down
and can't be changed.

this makes sure that you can't just take a 6-pipe "desktop" processor
WITH ECC and put it on a server motherboard. oh yes, that's also why
they have so stupidly many different pinouts. 

Imagine if a 6L6 rated for "audio" had a different base on it
than a 6L6 rated for "RF" use, and the socket for "RF" and
the RF-rated tube cost 5 times as much.  How would that have gone over?

that's the crap Intel is pulling each and every day.

Do you think that may be why a small number of huge users of
hardware have taken out architecture licenses and are starting
to fab their own server silicon? Maybe they don't like being the target
of extortion.

   -mo