
* check which functions are now in seriation
  and remove such from cba.

* move sdists and friends to a seperate
  package.

ceeboo 2009

* complete this list with known issues.

* improve the documentation.

* consider an option for handling of ties
  in clusterers (which support it).

* standard methods for ccfkms are missing

* unit tests to detect bugs caused by changes 
  in the R source code. at least my trust is
  gone with the copy on write issue.

* reconsider overloading of image.matrix

* replace Calloc with calloc, etc. and use
  own checks and cleanups.

* replace initializations of R vector memory
  that use a for loop with, e.g.
  memset(REAL(x), sizeof(double) * n)

* dist and sdist could issue a warning if
  NAs occur in the result.

* check again the duplicate or not issue :-(
  [current experience indicates it should
   work]

* recheck the handling of NA and NaN in dists
  and elsewhere.

* check if the R interface to ccfkms checks for
  unique initial solutions. reconsider how these
  solutions are generated. consider providing
  a default interface that searches for a "proper"
  number of clusters.

ceeboo 2006

+ rethink the current approach to dists in favor
  of the dist C level interface. this might be
  more developper friendly in the long run.

+ we cannot optionally include package Matrix.

+ the fixation on double for dists computations 
  may be too restrictive. at least we need to
  shift more checking to the C level functions due
  to the migration to package proxy.

ceeboo 2007





