Attendees - IG, JK, EB, PC, PM, GW

This meeting is intended to cover in detail the strengths and weaknesses of ROOT, and its deficiencies per the Run II requirements. Its basis is the CDF evaluation of ROOT wrt the PASFRG and PASSUMA requirements.

Irwin asked if there is an interface to DSPACK. Pasha stated that this is already done. Gordon stated that this is not that useful anyway. The question was raised about interface between the D0 object model and ROOT? Irwin stated that that's analysis, not data access, and thus would be covered later.

Gordon stated that when he used ROOT, to extract data, he had to know structure of the entire tree beforehand. Pasha stated that ROOT provides the capability to read whole event, doesn't matter.

G - can debug shared libraries in NT VC++

Pasha said that the development of templates is in progress. Berkeley is now merging CLHEP w/ ROOT. As for the D0OM interface, Gordon stated that if templates are in place, then there is hope for a D0OM interface. A converter is available now. Partial support is not good enough. Irwin said that nothing stopping it per se. Gordon said that this is not show stopper for September or even for December.

Multiple input streams are available (KS routine), operations on histograms.

CINT is great for prototyping - simplest constructs are there. CINT code can be plugged in and compiled directly.

No profiling of compiled or external code is available.

Modularity is not really there. Linking in pieces of analysis code to external is hard to do. Can a ROOT widget be placed into Python plot? Pasha said that this would be orthogonal to ROOT philosophy of complete integration of components. Irwin stated that we really would like both - whole packages, and the capability to use pieces. ROOT is basically intended to be a self-contained environment. Philippe said that a histogram widget would be easy to write. At present a histogram connected to where it's drawn.

The much-discussed ROOT robustness problem is mostly in CINT. User contribution is picking up.

Makefiles are appearing. BNL has gnu-build kit for ROOT.

Has its own memory control system.

Pasha raised the point that most of the complaints being raised about CINT have to do with the support and maintenance standpoint. From a user standpoint, it is a good product.

Three big issues remain wrt ROOT:

CINT

Modularity (can't use pieces)

D0OM interface