>You're a poster child for Dunning-Kruger

Actually I am an engineer who has designed many error correction circuits for communication and storage systems. I think I know how much I know about error correction systems, which is plenty for this conversation.

While the statement was made in Slashdot jackass style, the question is legitimate. Why didn't they do any or more ECC on the flash that is failing. There is probably a perfectly fine answer like "We knew the expected error rate and It was designed to last 10 times longer than the system", but the system lasted 40 time longer so the ECC correction capacity was exceeded". or "We had TMR, but the give then age of the system the error rate is now such that the error collision probability is too high". or "This flash claimed to be rad hardened but it turns out it isn;t".

I'd like to know the answer because I like techy shit.

The science indicates that your dogma is BS.

>You can detect single bit errors with a simple parity bit

You can detect (2^32-1)/(2^32) of every possible failure pattern with a CRC. With a combination of a multiple bit error correction algorithm (with most correction schemes n bits can be corrected with 2n redundant error correction bits) and then the CRC can be used to tell if you correctly corrected the data.

If you're so smart, why aren't you advocating using BCH codes or Reed Solomon codes or some form of forward error correction code over code and data stored in flash so random bit errors in flash won't affect the code that is stored in the flash? What is your super clever alternative?
 

They didn't do any ECC on the flash memory? I thought these people were rocket scientists.

Yes i'm from Great Britain, but what I think I that the colonials should be using 220-240V, not 110-120V. Since P=i^2*r and i is proportional to v, resistive power loss in the cable the cable would be cut to 1/4 of what it is today, greatly reducing the risk of excess heat in skinny power cables.

Any app writer can include their own magic number in the instance on the device and use that for pairing.

A topologist and an engineer walked into a bar.
The engineer kicked the shit out of the topologist for using the same words to mean different things than engineers use them to mean.

So read the rest of the post. FFS.

Solar cells on every house is great as long as there is local storage in every house too.

Wind power is great as long as there is good power distribution infrastructure: It's always blowing somewhere.

Nuclear power is great as long as you address operational safety and waste storage, both of which are addressable if you do engineering rather than politics. Part of that is again, good infrastructure so you can build the nukes in good places for nukes.

It's easy to point at any single generation or harvesting technology and identify it's flaws as a sole solution. However there are many technologies and combined together they form a robust and comparatively clean solutions.

If there was any data to suggest the ACT tests are statistically valid (they test the thing you think they test) or reliable (they would get the same result if you tested again) then the correlation may be a clue to something. However when the underlying test is neither valid nor reliable, the correlation it shows doesn't even show you there is correlation.

Layout tools, Schematic capture, logic simulators, analog and mixed signal simulators, P&R, floorplanning etc, etc.
The all have a GUI that needs to be used.

What's notable is that with all these tools, the specific ones I use in the company I work for making big-ass chips, precisely none of them work on a windows desktop. You either run them locally or remotely on a Linux desktop. As time goes on they tend to drop support for older unixes. I don't know anyone who runs these on anything except Linux these days and windows is just a platform to run X or VNC to get to the desktop of the Linux box running the tools.

Well if it does, they can just give up and go to the baker on foot.

1) Screw software, hardware is where it's at.

2) Hard topics pay well: DSP, information theory, crypto, information coding, etc.

Yes. They have good bread and public transit.