[optical-networks] OFC Paper Intro Material

[Joe Touch]

Joe Bannister wrote:

> Hi,
> 
> I've been trying to condense the "networking" part of the OFC paper into a
> crisp, formal description of the problem.  Here is my attempt at getting it
> down to one paragraph.  The actual work is being done by Joe T, i.e. to see
> if the routing table admits the kind of structure we're seeking.  One part
> of the problem -- sustaining an acceptable hit rate -- could be difficult to
> solve.


FWIW -

Correlators can overlap IF they map to the same output port.

e.g.:

		1xx00 = 	10000
				10100
				11000*
				11100

		x110x = 	01100
				01101
				11100*
				11101

Note that there is no more compact representation that uses one 
correlator with 3 or less target bits.

Further, and more specifically:

Correlators can overlap IFF there is an ordering in which matches to 
previous correlators override later ones, AND in which there are no 
false positives in given that prioritization.

Joe