formal schedule p3e or whatever

nightly build

studies for detector optimization
  looking at global parameters
  results depends on suble parts of algorithm
  may not be ready for some years

multi-track events as a goal

data format: raw vs. hddm

geometry spec.
  c++ interface
  ease of use
  documentation
  split it off

64-bit builds: David

Documentation discussion
  eugene mentioned

speeding up tracking

escaping local minima

formalizing/documenting basic classes used in reconstruction

locking tagged releases: subversion upgrade
  waiting on new webserver: halldweb1.jlab.org

design phase should be there

===from action items===

coding conventions: CLEO model?

===from brainstorming list===

alternate tracker

===from email lists===

structure for storing MC results

JLab farm issues

new platforms: 
 centos
 fedora
 ubuntu