POOL Historical Notes POOL has been the most

POOL Historical Notes • POOL has been the most advanced and the most used AA project. • Currently, excellent teamwork with experiments on new features and debugging • Poor communication about requirements early on between POOL team and experiments. For example, CMS assumed existing ROOT functionality (e. g. updatability, transient data members, embedded C++ class pointers) would be provided initially.

POOL concerns • POOL must demonstrate ROOT schema evolution for ROOT storage manager (in a few months, not years). Work with ROOT team. • work closely with experiments on POOL/ROOT optimization and tuning • implement missing features (e. g. Vincenzo’s talk, Atlas needs) • Document ROOT features that are not (yet) supported (e. g. variable dimensions of arrays). Also, document things that should work, but don’t (e. g. top level std: : vector<T> in a TTree)

POOL concerns (cont. ) • authentication and authorization in LCG file catalogs – don’t wait for magic solution provided by GRID tools. The File Catalog WP should provide authorization hooks. • more realistic test coverage – many bugs found by CMS are not reproducible in POOL tests – in particular multifile, multitransaction, multithread, etc. Incorporate use cases from data models of experiments. • user documentation!!!!!!! The only current integration guide is the examples and tests and the code itself.

POOL concerns (cont. ) • Browsability of My. SQL collections • Use of collections in parallel environment • Interactive ROOT browsing of persistent objects – Must be able to generate ROOT dictionaries to enable full browsability. – Ultimate solution – unified dictionaries (SEAL)