Summary of July 31 PWG Software Status Review Session
- Stability of the software system continues to improve.
- A revised protocol for dev -> new migration has been defined based on recently-established new QA diagnostics. This protocol will enable greater usage of a stable, QAd, up-to-date new library. Infra development should occur in .dev, algorithm development in dev, tracking parameter and physics optimizations in new and bulk production in pro.
- Persistent StEvent stability and functionality are now critical issues. The new version of StEvent which includes the revised DST format will be posted and described in detail before release to allow appropriate discussions.
- A notification/execution sequence is required for code/library changes so that users can better schedule their activities. Notification of changes should be placed on a commonly-used list server with adequate notice (at least 24 hours).
- All code developers should heed established CVS etiquette, including adequate documentation of code changes in the CVS repository.
- A new software package proposal mechanism will be put in place soon. Each new package will be reviewed before being stored in the official repository. The intention is to keep the process rather unbureaucratic.
- Gene Van Buren has been named Software Documentation Coordinator.
- Documentation enforcement should be based on peer pressure. Documentation status of code elements as evaluated by Gene should be prominently displayed on the web.
- Documentation of general-purpose code and infrastructure will be organized directly by Gene.
- html-based web documentation is strongly preferred.
- A prioritized list of the most essential missing documents will be prepared by Gene and posted to the web.
- A ROOT `folklore' resource is needed. Gene will organize this.
SEG (Software Executive Group: Jacobs, Pruneau, Ullrich, Wenaus)
- Input to SEG comes from all software workers.
- SEG is in part a strategic planning group and in part a conduit for the inevitable grievances, frustrations and software community problems that require resolution by negotiation among a few principals.
- SEG's weekly activities are posted to the web as a matter of record.
Multiple Data Streams
Positive news: The buffer box will be augmented to deal with up to 15 event-type streams. Negotiation is still required with RCF to maintain reasonable event-type contiguity in HPSS.
- Dave Zimmermann will move to BNL on Sept. 1 to commence integration of GC software into the STAR/RCF infrastructure.
- A workshop will be held around Oct. 1 to discuss GC deliverables with the user community.
- Greater coordination is required among PWGs and GC personnel on protocols for post-DST analysis and use of GC and other infrastructure on CAS. miniDST and uDST design and usage are important open questions.
- Data-base development is coordinated and supervised by Jeff Porter.
- Calibrations DB system is of central and immediate importance. This system is on the critical path and needs much input and development from the PWGs. An immediate issue is the transition from parameter files to a formal DB with flexible key system.
- Offline and Online calibration procedures and algorithms are needed.
- Major initiatives from the PWGs are required, which are also good entry-level tasks for new participants.
- Calibration group activities must now be expanded to involve the whole PWG community.
- Target date for implementation of the TPC DB is ~Aug. 20.
- With DBs in place all algorithm parameters should be extracted from code or parameter files and placed in appropriate DBs to factor the code optimization problem into algorithm development (CVS versions) and parameter tuning (DB keyed versions).
- A formal procedure is required to insure that GSTAR contains up-to-date representations of the physical STAR detector in its various configurations. The STAR detector support group (Bill Christie) should collect periodic subsystem updates from subsystem representatives or project leaders and communicate these to Pavel for incorporation into GSTAR.
- Simulations QA should be brought into line with recent developments in software/infrastructure QA. A formal QA process is required prior to bulk production of simulations data.
- The PWG prioritized event list requires update. Frequent feedback via webpage documentation is needed on simulations production status.
Much progress has been made in recent months in establishing a QA data base of histograms and other elements. Procedures have been outlined to regulate library migrations. More detailed criteria are needed for DST event and run rejection based on quality. Software QA procedures should be adopted also for simulations QA. DST file types and self documentation Self-documentation, stability and backward compatibility of .root files are still open issues. At least until these questions are resolved all DST files should be written in .xdf format as well as .root format.
These bi-weekly meetings have been of central importance for rapid development of the STAR physics program and code optimization. The meetings should continue with the existing schedule until the resumption of RHIC operations in December, when we may review the schedule.