From a correspondence with KeithLofstrom ...
Keith -- Here is a position paper I wrote for a workshop. The idea is that addressing is so ubiquitous that it could be blinding us to other alternatives.
The cybords project hadn't been hatched yet when I wrote that. Now on cybords I'm trying to explore the alternative space a little bit. For example, cybords don't have addresses. Messages have a channel number and boards can listen to what ever channels they want. The cynase protocol looks more like a radio protocol than a buss protocol.
Here is the plan so far:
Loss and corruption is frequent but not dominant. The controllers' statistical approach to signaling make them tolerant of many mistakes, including programming mistakes it turns out.
- Each board connects on four sides.
- Opposite sites are electrically connected making two orthogonal busses
- Controllers relay messages between busses
- Signaling is async bit serial. Collisions are possible.
- Designs must be tolerant of lost and corrupted messages
So I am thinking of bussing together the cards in unusual ways. That is what I'd like for you to "blue sky" about. Here is what I've considered so far ...
The third alternative is attractive to me because two sufficiently distant nodes could communicate with their near neighbors in what would be a corrupting collision in the other two schemes. This is a sort of "capture" effect. My brother (who got me started fooling with the AVR parts) thinks that this is dangerous because intermediate nodes would see non-digital signals during a collision. He suggests I at least study the pin drivers, which are available, so that I understand what would be happening electrically in and after a collision.
- Pull down against a (possibly distributed) pullup resistor.
- Pull up and down through a (short circuit protecting) resistor.
- Pull up and down a buss segment which is isolated by series resistors.
I should point out that high bandwidth utilization is not a priority for my designs. My plan is to throw away a couple of orders of magnitude in order to get unusual connectivity properties. I'm also aware of some of the more modern serial protocols but don't see how they advance my philosophical goals.
Thanks for your attention. Happy holidays. See you soon. -- Ward