Back to POOL Home page

                           Minutes of the POOL meeting, CERN, 24.08.04
General News 

Proposed Authorlist for the CHEP contributions agreed (with some corrections after the meeting):

Kristo Karr, David Malon, Alexandre Vaniachine
(Argonne National Laboratory, Argonne, IL 60439, USA)

Radovan Chytracek*, Dirk Duellmann, Markus Frank,
Maria Girone*, Giacomo Govi*, Jakub Moscicki*,
Ioannis Papadopoulos, Helmut Schmuecker
(CERN, 1211 Geneve 23, Switzerland, *PPARC funding)

Zhen Xie
(Princeton University, Princeton, NJ 08544, USA )

Tim Barrass
(University of Bristol, BS8 1TL, UK)

Carmine Cioffi
(University of Oxford, Oxford, OX13NP, UK)

William Tanenbaum
(Fermi National Accelerator Laboratory, Batavia, IL 60510, USA)

For the experiment integration and DC experienece talks we expect to extend this list with the appropriate experiment people provinding input to these contributions.

WP Coordinators reports 

- Markus: LHCb verification of POOL 2.0 builds shows still problems
and makes them not usable. We should make sure (eg in the release notes)
that people are aware of the fact that our ROOT4 (POOL2) builds do not
have a frozen file format yet and that data written now with the pre-releases
may not be readable later.
Reading old root3 files with POOL 2-pre is of course not affected and even
updating them did work in Markus' tests, but root4 new files needs agreement
with the ROOT team in particular for STL stograge of containers which ROOT4
does not support yet (but later). Markus would like to avoid another
intermediate streaming format which needs to be migrated once root4 comes
up with direct support later. It would be easiest to stay to the old root3
streaming in pool to limit the number of formats to support in the longer
term but this may limit the schema evolution suport in POOL2 for these
STL containers.

For this reason Markus did advice ATLAS to carefully check if the still
unvalidated POOL2 builds could be picked up for the ATLAS DC.

Markus has in the meantime started the discussion with Phillipe Canal about
future data formats and will report in the next meeting about the outcome.
Atlas should be kept informed about any development there.

- Markus//Helmut/Ioannis with help of Rene & Andreas
POOL2 proper release will need a new root version because of some
problems discovered by Helmut on the root side with the cint handling
of precompiled STL headers. Rene has been helping and the root team has
tagged root 4.00.8b with the required fixes which we should pick up
ASAP in a SPI provided toolbox. Andreas has already started to work
on the root build for that.

- Ioannis: minor interface changes in RAL have been introduced
to bind the type conversion to a single session rather than having
it shared for all db sessions. This allows to use a different SQL to
C++ mapping in one application eg for different data domains.

Progress on using RAL plugins (eg odbc) for serving more
than one db backend (eg mysql and oracle)
Rado - odbc plugin should come this week (revised later next week)

Progress also in the implementation of the object relational storage:
work split xml mapping import (Giacomo), construction of default
mapping from headers (Zhen), basic object io (Ioannis)
Should soon be able to excercise the full chsin

- Markus: Pool 64 bit port has been made work within Openlab student and
passes now the POOL tests. Need to integrate the (inline) code
changes (typically long -> int) before the head version diverges
too much from the one used in the openlab.

Zhen: two bug fixes in composite catalogs
- will propose interface split into basic lookup and meta data part
by end of week
- communicate after agreement in pool to globus / egee / gd
- first implementation can't be before end of october because
of vacation, csc and chep.

Helmut: collection catalog prototype with significant functionality
has been produced.
-Needs to be tested by experiments.

David M.: may not be used by ATLAS who are also looking at AMI
for this purpose but could serve as reference implementation.
Biggest problem in this area: unclear how this functionality
will be used by the experiments

- Giacomo: last SPI toolbox allows to integrate Qmtest
- POOL 1.8 will pick this up and complete the milestone

- RD pointed out that ATLAS is still waiting on feedback on an
open bug concerning the statistics output from POOL.
Markus could not yet reproduce the problem but will take a
a look at the detail provided by Peter v.G. in the bug report.

Next meeting usual in two weeks.