Tue Jun  3 12:52:42 PDT 2008
============================

  Fixed problem with filter-stage-1.prl reported by Gyorgy Abrusan.

Fri Feb  8 15:13:12 PST 2008
============================

  RepeatScout Changes: Robert Hubley, Institute for Systems Biology

     - Wrote new program to replace build_lmer_table ( called elmer ). 
       This program can handle full genomes for lmers up to 16bp.
     - Simplified forward/reverse lmer handling by removing the need
       for 2 copies of the input sequence in memory.  This change makes
       it easier to calculate and enforce sequence boundaries.
     - Modified RepeatScout to honor sequence boundaries:
           - Continues to load sequence(s) into single concatenated array.
           - Created new data structure to hold start positions of 
             each fasta header.
           - Uses seq boundaries in extend_right/extend_left to make
             sequences non-active if they run into a boundary.
           - Uses seq boundaries in maskextend_right/maskextend_left to
             prohibit reduction of frequency of lmers which span boundaries.


  WISHLIST:
     - Reduce the memory footprint:
        - Sequence is in bytes  ( 1 byte/base could be reduced to 3 bits/base )
        - build_all_pos() creates a linked list of all positions for all 
          frequent lmers ( default > 3 ) in the input sequence at the start
          of the program. This could/should be the function of
          build_lmer_table() and for large genomes can be cached to disk.
          This would require all RS functions to be able to read from a disk
          based hash.
        - master sequences ( consensus models ) are kept over the course
          of the program.  ( not necessary -- flush to disk ).
     - Support full custom matrices in addition to the Match/Mismatch based
       scoring system.