Attendees - Gaines, Canal, Kallenbach, Watts, Wyatt Merritt, Murat, Kennedy

Philippe presented the preliminary comparison of Nirvana to requirements documents. From ensuing discussion:

Functionality of note:

The product uses IPC for remote snooping, and Minuit fitting. The capability exists to draw onto a histogram a function of the x-axis variable, not of arbitrary variable. Presentation-quality output is the only missing piece. A re-working of the postscript generation is in progress.

The point on the slide regarding a problem with data access permissions is irrelevant

Issues:

Pasha brought up the point that use of Nirvana will require users to learn Python, which they would rather not do. Wyatt stated that Python is the official D0 on-line language. The issues of a scripting language and api were raised. The requirement for COMIS/KUIP functionality was mentioned, and Gordon noted that Python would have that by default. It was also noted that by using Python for the scripting language, we would grab the existing Python user base.

API will provide object-oriented access to data and to the gui. There will be api/scripting interface, gui/scripting interface.

Other functionality missing at present:

Event mask (streaming) - select some events, read them back w/o writing out to separate store.

Command line unavailable

Save configuration in file using gui

The rest of the meeting was spent discussing ROOT.

Status:

Rob showed slides with the current status of the evaluation (updated from last week)

Rob is building the Linux KCC version, which is not distributed. ALICE (LHC) has made a commitment, along with NA49. Duke and the SVX2 test stand will be writing ROOT Ntuples.

Issues:

Rob stated that support from the ROOT team is excellent, but that a great deal of effort will be needed to fix the deficiencies (probably provided by us). The first round(s) of upgrades will be significant and we will want to get a lot of user feedback during this period. The point was raised that by choosing Nirvana or another local product, we would have complete control, and thus avoid FROOT.

The issue of the severe problems with CINT was discussed. In particular, it requires the user to learn what it not allowed, and that some trivial operations are not allowed. Rene's view of CINT is that it's there, and basically functional, and thus updating it is a low priority. Discussion of purchase of a commercial replacement took place. Wyatt said that $60,000 would be cheap compared to fixing up CINT or writing a replacement. Gordon said that such a product would really be only a front end, and not a replacement. It was agreed that interfacing to ROOT is messy regardless of the tool used.

Gordon stated some of his own observations of ROOT - he finds it less usable (e. g. memory management, error reporting) than Nirvana. He also finds that the GUI builder is not good, and that the trees were in such bad shape that his experiment had to be aborted. It is his opinion that starting with ROOT, investigating $60,000 and 2 man-years, the product would be no farther along than Nirvana with the same effort. He also said that no one at D0 is using ROOT presently.

Pasha presented some slides - comments on Gordon's report

Next: Friday - MATLAB, Octave status