Minutes of DFWG Meeting
Apr. 29, 2004

Present:  H. Greenlee, F. Deliot, S. Kulik, 
Video:    G. Watts, S. Protopopescu

Gordon did some speed tests on tmb_tree and top_tree, using supplied
headers and headers generated using MakeClass, for whole and partial
events.  The main conclusions are as follows (see agenda for detailed
results):

1.  Any root format is much faster than tmbs.

2.  Speed of reading root files depends mainly on the event size.
MakeClass is no different that presupplied headers.

3.  There are significant gains from reading a subset of the entire event.

4.  There is significant overhead from opening new files with TChain,
depending of course on the size of each file.

There are two new cvs packages called d0dfwg_reports and d0dfwg_tests
(thanks Gordon).  Frederic has already contributed his chapter on csg
to d0dfwg_reports.

Things to do
------------

Gordon will refine his timing tests in a more controlled way, and
using any additional format root tuples that other group members can
supply.  Gordon will also test higgs_multijet (a special purpose
format for large numbers of small events).

Slava will document top_tree and supply standard tmbs and top_trees.

Frederic will find some wz_analyze and qcd_analyze tuples.  Frederic
will also write a summary of other formats (wz_analyze, qcd_analyze,
Athena?).

Serban will document tmb_tree and produce standard tmb_trees from
standard tmbs.

Herb will organize e-mails to d0dfwg mailing list and provide skeleton
report.  Herb will also supply Athena root tuples for Gordon to test.

Adam will be prevailed on to write edmroot chapter.