The Markov give me queue sizes ina log3 N bi computer

 Have to leave a certain bunch of arrivals uncounted, in the queue.
Each time I increase the number of bits in my log3 countera larger error term in one of the  queue size grown.
I am minimizing transaction counts.  For each under sampling I incur error. The log3 computer tells me how many message I can out out if the error term is optimlly spread.

Every channel rate I count up y states.  up to some N tribis then stop, I can get no more accuate.

Then give me some large set of items, an I want to count them in three space.  I can derive the error term, and this should work for physics. Plank's constant is the error in counting Avagadros number.