I am too a paranoid about flash wear (in fact during write->compile->upload to tablet I sometimes unwittingly try to minimize the number of times I upload). But almost everyone told me that I shouldn't really worry.

On the builtin flash, it's done by the jffs2 driver.
On the internal/external cards, it's done by the card itself.

Yes the internal flash uses jffs2 filesystem which does wear leveling. Also SD cards and internal MMC in N810 does wear leveling too.

I would worry more about inability to recover data from internal flash when tablet does not boot anymore. Also jffs2 is quite slow. That would be my reasons for booting whole OS from SD card and not using internal flash.

thats one reason why i want to get my hands on a touchbook...

What a weird answer. I'm using my N800 nearly 2 years or so. Also there is several other n8x0 owners and they're fine. And screens are completely Ok. And why screen should die?

I'm only aware that on Nokia 770 there was hardware design error which caused LCD to die upon restart or power-up with some probability. But this error does not exists in n800 and n810 to the best of my knowledge so they survive reboots and power-on without prob's. Do you have any proof?

I don't know why, I suspect a bad batch or a design issue, mine definitely died right after one year of use. There are several reports on this forum of screen failing around the one year mark.

my usa market 770 died fairly soon after i got it (damn it expansys, why did you have to send me one of your leftover usa stock?!), but my second hand european N800 seems to keep on going and going...

Several out of ... how many devices? And at which time frame? Let's say if you will produce thousands of complicated electronic devices, I can almost guarantee that you will encounter some dead devices at QA stage and some device deaths at their life cycle. Just because they're complicated electronic devices and many other reasons.

Can I have simple math? Let's say, there is 5 failed devices. Sounds awful, right? But what if these were 5 out of 1000 devices within year? That's failure rate of 0.5% per year. Very good for complicated devices. And actually not awful but very great result. Surely, there will be 5 unhappy humans. And 995 happy ones. Overall result is good, then. And probably better than industry average.

So, "several reports" are not proof on it's own - AFAIK Nokia sold many and many thousands of these devices so it's ok to see several dead devices (just as several dead iPhones, blackberries, ... whatever). It is matter of statistic - how many reports out of how many device owners at which time frame and which chances are to became loser in this "game". For Nokia 770 failure rate was high enough and there were known methods to cause issue to appear with high chances - perform many reboots or power cycles. And there was good explanation of problem and it's details. So there were special threads about problems and quite many reports and details were known. For n8x0 I did not encountered such mass failure reports so far nor I seen problem analisys. On my own experience N8x0 are not dying quickly due to LCD faults (it's more likely to physically damage it, if device falls down, etc). That's why I'm a bit skeptical about such LCD posts.

Backing up critical data:

For over 10 years I have used an odd home-grown method.

I get a new computer every few years. It always has a way bigger disk system than the old one. I use GHOST to obtain a backup of the old computer ( I also do this every few weeks, backing up to three different external drives). The entire content of the old computer's GHOST backup goes into a folder called something like "oldcomputer".

After I get the new computer tweaked and happy, I go into the old computer ghost file and extract and copy relevant portions to the new computer's file system. Eventually the "oldcomputer" GHOST image is fully transferred to the new computer and I then delete it.

I keep one of my external backup drives at work, one in my car, and one at home in a fireproof safe.

I also use Mozy for on-line backup.

I have pretty much every file I've ever created since ~ 1992, in five different places.

Steve

Oh, sure, but what if those reporting the issue are only 10% of those having it? Suddenly your 0.5% goes to 5% which is pretty bad.
I don't know how many tablets nokia sold (they never disclosed the number), much less the number of failed screens (since I replaced mine myself, mainly because nokia didn't reply to my request for help, it wouldn't show up in nokia's statistics), but it's not very far-fetched assuming that the members of this forum are 10% (or even less) of all tablet users.
Going back to the original topic, due to the stated characteristics of nand flash (which I don't know if I should believe) it's more probable than something else (including the screen) will fail before the memory.

I have a simple rule - don't trust any single media or technology But seriously, the physical longevity is just one factor and it's often overlooked that backup is a *procedure* not just a bunch of data on a supposedly safe media. It doesn't matter if a DVD lasts for a 1000 years if I drop it face down and lose the data because of bad handling.

So, in my book, something I'd really call backup (and not just a spare copy) is something that exists at least on two different types of media in at least two different physical locations. DVD's are too small and too slow these days, had bad experience with RWs and I could make a landfill if I used Rs, BR is not really an option where I live ($$$), so that gives: 1TB NAS backup with the really important data periodically compressed/crypted/synched to Amazon S3.

PS. On flash wear: A while back I managed to run down a 512MB flash in a little over 6 months in 24/7 conditions (was used as a swap partition in a NSLU2, long story). I don't really see that as a potential problem in any real-life scenarios, though (you're not running your NIT swapping 24/7, are you ?).