There is still work to be done on the cluster/hitfinding software. Examples of such work follow.
To move towards more realistic software that can function outside idealized simulations, a way of handling missing or ``hot'' electronics channels must be implemented. Similarly, a way to correct for field distortions at the hitfinder level should be implemented. Also, a ``front end'' must be incorporated that will unpack the 8-bit ADC data (compressed from the 10-bit data by ASICs in the front end electronics), and possibly apply offline pedestal, gain, and time-offset corrections. This will likely be implemented as another analysis package that precedes TCL.
It will be worthwhile to compare quantitatively the offline cluster/hitfinder (described here), and the faster one used in the trigger software. Such a comparison is planned.
Currently, there are not too many switches to tune in TCL (a desirable situation), but for those that do affect performance (e.g. TCLPAR.MF_PV), it is neccessary to see which setting works best with the tracking software that follows TCL.
It is often envisioned that, for analyses like HBT where resolution of overlapping hits may be important, a ``second pass'' at hitfinding may be made, using tracking information obtained in the first pass. Implementation of this mode of running does not exist yet.
Estimating the crossing angle of the parent track from the shape of the cluster pattern is an interesting and possibly useful twist to clusterfinding. As discussed above, our understanding of this estimation could be improved.