Summary of ideas coming out of BNL meeting with Gunther, Inkyu, Heinz, and Steve 6/9/99

• Cut on dE/dx to select min-I tracks in field off sample.   This should minimize the multiple scattering.
• This means Inkyu's TPhTools IsFindable routine probably gives a reasonable starting track seed (from the vector going from the IP to the first hit in layer 1 or 2).
• Expect main hardware position variations in the local sensor x and y.  The other variation one might expect will be a tilt or rotation around the axis of support (sensors are held at the top on one edge).
• There is a member function that swims a track to a sensor (... try TPhSwim, SwimTrack)
• Relate reconstructed and MC hits via location on sensor. 
• Relate reconstructed and MC tracks via number of shared hits.   Gunther is planning on working on this.  Inkyu has an unofficial function that does this now ... TPhTools.GetNHits and TPhTools.GetNCommonHits
• One should be able to use the multiple vertex positions about the beam orbit to really absolutely nail down the positions of the first two layers with respect to the vertex with field off data.  Global algorithm:  allow first two layers freedom to move and nail down position from multiple vertices.  Once these positions are fixed, use the vertex and first two layers as a seed and work your way out aligning each layer with respect to the growing seed.  In the back of the horseshoe there is a segment of the detector that is not illuminated by straight line tracks passing through the front of the detector, regardless of vertex postion.  Align this part in the same way you align the first two planes of the detector.  So, global algorithm will have two seeds at least.
• Tracks passing near the interface between two pads are very special and have a much better position resolution (maybe 20-30 microns) than tracks passing thru other sections of the detector.  This position information may be very useful.