ok i'm imagining it in my head
   a's are a column of dots on the left with one color
   b's are a column of dots on the right with another color
   there are only a few b's, and lots of a's
   "some b's are a's"
    you can look and see that b's that are near a's, are more likely to be
   them.
   we can abstract that nearness metric away to provide for overhead
   statistics later.
   then all we need is properties among the b's. any additional associated
   data with each packet.