Lynn,

all you need is to have programmable fuses (best if they are EEPROM). Then the user can choose the polarity, duration etc. of reset during programming, all you need to do is to provide enough options (and determine some reasonable default, especially if the primary programming method is going to be ISP).

Just look at some of the ATMega datasheets, for inspiration.

Please consider also the internal RC oscillator, and an early-warning power-failure comparator/interrupt (a.k.a. "BOD"). You might also perhaps spend some time rereading my "reset" document and commenting on items there - this is why I wrote it, to gather the ideas related to the topic.

JW



PS.
1. In EVERY design with a microprocessor/microcontroller we ever made, we used a proper reset IC, whether it was 7705, 1232, 707 or 69x - and we continue to do so even with chips which claim internal reset, simply because their datasheets give little guarantee in that respect. With such IC, it is almost impossible to determine exactly the startup time, ergo, we don't really care about a millisecond or two.

2. The mentioned supervisor we use would reset the chip under the same conditions anyway, so that's no difference.

3. Negative reset ICs are easier to obtain both as OC and push-pull; and Richard likes them more, too :-) Of course this is not an issue if you can give a guarantee that the internal reset will work no matter what (or at least the same guarantee the usage of a standard reset IC gives).