First, not to confuse everybody, I am not promoting a mcu-based programmer as the first step; it was just an example how a "burner" - and IMHO a better one than the LPT-based constructions - can and does fail, for trivial (but for newbies most probably restrictive) reasons.

Throughout this whole thread, I am still only thinking out "aloud", I still can't say this or that is a perfect solution (if there would be such).

The original premise is, that the newbies can't get the original Atmel cable or the 74VHC chips needed, and the cost is not the only reason.

Second thought is, that even with the original Atmel cable and the original Atmel software, most of the likely reasons for the programming to fail (as I listed them in my original posts and restated a zillion of times after) still persist and it is not likely to have an easy remedy (or detection) with the existing setup.

So, what I want to do, is
- cheaper
- more accessible (using commonly available parts)
- supportible (original design, possibly open source in all respect, with concise explanation)
- better (having extra features to avoid common problems)

Jan Waclawek