Minutes of the
                     Jefferson Lab CLAS Collaboration 
                           January (20,21,22), 2000

Next meeting: TIME NOT YET DEFINED, see point 4, below.

For you to do:

1) Check out the final "as approved" revised shift bylaw, which is
   included below.

2) Check out the final "as approved" publications bylaw, which is
   included below.

3) If your talent permits, help in the effort to benchmark the speed,
   efficiency, and resolution of the first-pass analysis code.  Help
   optimize the code, such that the time spent cooking is again less than
   the time spent taking data!

4) The dates we selected for the next meeting, May 18-20, will not
   work due to JLab schedule conflicts.  Possible alternate dates are 
   May 4-6 (Thursday thru Saturday) and May 14-16 (Sunday thru Tuesday).
   If you have an opinion about which dates are good, bad, or impossible, 
   please send me a brief email at schumacher+@cmu.edu.
________________________________________________________________________


	Not even a Virginia snowstorm that delayed the opening of the
lab by two hours could prevent the CLAS physicists from starting their
winter collaboration meeting promptly at 9 am on a cold, slushy, dark
morning.  In total, 92 collaboration members signed in to the meeting
over a period of three days.

	Data-taking with CLAS in 1999 was enormously successful, with
16 BILLION triggers written to tape, compared to about 5 billion in
1998.  This resulted from both good performance by the accelerator and
the smooth and reliable operation of the CLAS spectrometer.  In
particular, the last of the DAQ speed problems were solved, such that
routine operations became possible above 2,500 recorded triggers/sec.
Averaged over calendar days, this amounted to about 450 million
triggers per day.  Present rates of data cooking range in the
neighborhood of 80 million triggers per day.  Hence there is presently
a mismatch in rates, by a factor of 6, leading to a growing backlog of
data.  The exact factor and how one chooses to average the numbers can
be debated, but the collaboration recognized that this is a serious
present problem which needs to be addressed.  The causes and solution
to this problem were one of the main topics of discussion at this
meeting.  It was introduced by Bernhard Mecking at the beginning of
the first day and discussed in more detail on the second day.

	It is likely that a new NSAC long-range planning exercise will
get underway this fall.  Bernhard Mecking outlined some thinking about
how CLAS could fit into the long-range plan.  In particular, CLAS
should maintain its ability to identify exclusive electron-scattering
processes at higher energies, while real photon experiments may
instead be concentrated in Hall D.  To this end, there are ideas for
replacing the inner tracking region of CLAS with (new) full-coverage
tracking, plus more hermetic electromagnetic calorimetry and possible
RICH Cerenkov light detection.  Some prototype work is underway (but
was not presented at this meeting).

	The overall status of the analysis projects in CLAS was
reviewed by Reinhard Schumacher.  There are presently 67 analysis
projects on the books of which 55 have data from CLAS to analyze.
There are 36 PhD students working on CLAS data.  Of the 55 projects
with data to analyze there are 19 with no students, but many of these
are in the hands of post-doctoral or permanent staff researchers.
Only 9 analyses are inactive at this point.  The intermediate results
of 15 analysis projects have been shown at conferences, both APS and
Topical.  It appears that as many as three analysis projects can
evolve into publication submissions by the middle of the year.  It
should be noted that CLAS started taking production data early in
1998; if we can keep up the present pace then first publications will
have been submitted two and a half years later, which is not atypical
for any large complicated experiment (or even a simple experiment).

	Most of Thursday was devoted to a "Physics Fest" of results
and analysis reports; see the agenda sent prior to the meeting.  It
would not be possible to give a full and fair report on all of them
here.  Notable were three presentations of nearly-final results: 1)
Steve Dytman reported on eta electroproduction in the region of the
S11(1535) resonance, results based on the work mainly of Richard
Thompson. He estimated that the cross sections systematic
uncertainties were under control at the level of 8%, which was
comparable to or better than the statistical uncertainties.
Systematic effects such as the dependence of the results on the choice
of event generator were discussed. There were hints in the data of
interference between the S11 with the D13 resonance, and quite
precise-looking values for the interference structure functions
sigmaTT and sigmaLT were shown.  2) Cole Smith reported on pi0
electroproduction, with the goal of extracting the small amplitudes in
Delta excitation.  Many systematic effects have been studied, such as
the sensitivity of the results to the cuts, the event generator, and
the radiative corrections.  Comparison was made to the MAID
parameterizations as a consistency check of the CLAS data with the
rest of the world.  The goal of extracting, for example, the ratios of
the E1+/M1+ and S1+/M1+ amplitudes, was seen to be close at hand.
Preliminary results were shown.  3) Claude Marchand showed the results
for phi photoproduction at high t, plus a few related g6 analysis
topics.  Background subtractions and other systematic effects were
discussed.  Phi angular decay data were shown for the first time, as
was further progress in the analysis of rho and omega photoproduction
at high t.  The phi cross-section data, which probe the reaction
mechanism of Pomeron and 2-gluon exchange, were declared to be in
publishable shape.

	Paul Stoler reported on the recent 12 GeV Physics Workshop
held at Jefferson Lab.  The CLAS upgrade which is envisioned would
support a program of (i) exclusive meson electroproduction at high W
and Q2 but at low t: this program would seek to probe the off-forward
parton distributions; and (ii) running at low W but still high Q2 and
high t: this would probe baryon form-factors out to larger values of
momentum transfer.  The comparative advantage of CLAS over Compass and
Hermes would be the ability to identify with good resolution various
exclusive final states.  A program of physics needs to be assembled in
time for the summer PAC meeting.  Paul Stoler will be coordinating
this effort.

	A long series of short talks itemized the progress which the 8
active run groups have made recently in the calibration and cooking of
their data.  A common theme was that the access to the JLab data silo
is presently very inefficient, and that the process of cooking is
painfully slow.  At least one run group proposed to take the data for
its running period off site, after setting up a separate farm and silo
system at a different place.

	Dieter Cords presented results of work on the on-line system.
The Level 2 Trigger is now in routine use, thanks to the hard work of
Dave Doughty and associates.  During the recent g3 run (photons on 3He
and 4He targets) the Level 1 trigger rate was 8.6 kHz, while the Level
2 acceptance rate was 2.2 kHz.  Together with a prescaled and neutral
triggers, the total rate to disk was an astonishing 3.2 kHz.

	Mark Ito then discussed the situation in the offline analysis.
For a long time, the farm of computers that "cooks" the CLAS raw data
(to find the tracks and apply the myriad calibrations) has been idle
nearly 50% of the time.  This was due to extremely inefficient
queueing of data tapes in the silo.  Mark has worked with the Computer
Center and several collaborators to organize a more efficient
silo-access scheme.  This includes new restrictions on requesting data
from tapes by jobs that are already occupying farm nodes.  Also, up to
10 new tapedrives will be on line soon, albeit of a different type
than our raw data tapes.  There are signs that the new scheme is
already working.  A factor in through-put of up to 2 can be obtained,
in principle, which is not enough to solve our problems.

	The speed of the analysis codes are such that we cannot cook
data at the same rate that it is produced.  While this is not an
unusual state of affairs in physics experiments, for CLAS it is a
problem as long as the detector continues to work so well that the
flow of data continues to build a backlog of uncooked data.  Mark Ito
cited anecdotal evidence that between PROD1-2 and PROD1-9 the code
slowed by a factor of 2 to 3.  [Franz Klein has done some interesting
benchmarking, but more work is needed.]  Cooking the 16 billion
triggers from 1999 at a rate of 25 million triggers per day (a
recently-seen "good" effective cooking rate, before silo problems
started to be addressed) would take 1.8 years.  Later in the meeting
Jim Mueller emphasized the iron triangle of code performance: all code
has some efficiency, some resolution, and some speed.  Changing one
generally changes the other two.  We face the urgent need to better
characterize the analysis code we are using and to optimize its
performance.  For people who can help, no detailed knowledge of the
code is not even required, since non-specialists can still process
test runs that are needed to test various options we could use.  The
relevant time scale on which we need to solve the problem is 3.5
months, since that is when both e1b and e2 are estimating completion
of their present cooking.  A number of people stepped forward to
participate, including Mac Mestayer, Franz Klein, and at least two
others.  If you feel you can contribute to this crucial cause, please
contact Mac Mestayer.

	A group of collaborators are working on a database scheme to
control input and output to the "map" of calibration constant used by
all the analyses.  It would solve the problem of having analysis
results change unexpectedly due to changes in constants (often made by
unknown parties) by maintaining a history of map changes.  This system
may be ready for use by the time of the next collaboration meeting.

	A data summary tape (DST) format for the eg1 run group was
presented by T. Forest as a model for what other run groups may want
to do to cut down data files to a manageable size for physics analysis.
The eg1 model provides a compression factor of 71 by selectively
keeping information and packing it with FPACK.  Events are packed in
groups (according to helicity bucket) with reduced header information
for each event.  Conversion from cooked file to DST is very fast, on
the order of 7 minutes per cooked data file.  The eg1 Run Group has
the goal of putting its entire data set on a smallish RAID array in
its DST format.

	Will Brooks and Gabriel Niculescu reported progress with GSIM
acceptance studies.  Using elastic_gen --> GSIM --> gpp -->
reconstruction, they find results in cross sections that are "within
5% agreement" with world data.  Angles smaller than 20 degrees still
present problems.  Steve Barrow showed how scattering of hadrons from
target chamber ribs can cause significant tails in physics
distributions due to bin migration.  Cutting events at the
event-generator level because they are not heading into the fiducial
volume can be dangerous.  K.-S. Joo reported on the effects of
iterating the radiative corrections applied to the data based on
feeding observed event distributions back into the event generator.
He observed significant effects in at least some kinematic regimes
upon going a second iteration.

	Tracking efficiency was discussed by Jim Mueller, who reported
interesting results using events of the type e p --> e' pi+ pi- (p),
where the (p) was not found directly in the track reconstruction, but
which was known to be there from missing mass analysis.  Using a
well-defined set of such event he showed that the PROD1-3 code has
large forward-angle inefficiency for low momentum protons and pions,
while the PROD1-9 code has lower but still annoying angle-dependent
inefficiency for low momentum protons.  He traced this to the fact
that the values for the drift chamber resolutions (the DOCA-dependent
residuals) are not realistic.  When he widened these values in the
map, suddenly the forward-angle low-momentum resolution became much
better.  [This breakthrough is very new, and Jim is still studying it,
but it seems that he has found an important handle for tuning the
code.]  He also produced a efficiency function to use with the
existing codes which gives the single-track efficiency as a function
of momentum and angle for positive hadrons.

	On Saturday the Chairpersons of the Physics Working Groups
gave summaries of the PWG meetings; we will not summarize the
summaries here.  Also, Volker Burkert and Will Brooks gave overviews
of plans for the upcoming e1c and e5 runs, respectively.  Notable was
that the dual H2/D2 target for e5 is now built and undergoing tests.

	Membership committee chair Gail Dodge reported that three
persons were discussed by the committee and recommended for full
membership to the collaboration: James Kellie and Ken Livingston from
the University of Glasgow, and Gerry Feldman from George Washington
University.  Their memberships will be voted on at the next
collaboration meeting.  A vote was taken on the full membership of
Daniel Carman of Ohio University, and he was accepted.

	Hall Crannell spoke for the Service Work Committee.
Discussion reaffirmed the value of the work that this committee is
doing for the collaboration.  There were no major outstanding issues
at this time.

	Volker Burkert spoke on behalf of the Speaker's Committee.  A
number of invitations for talks have been received and approved at
various conferences [see Brian Raue's web page], but none (yet) at the
Intersections Conference.

	A several-hour discussion occurred in debating two bylaws, the
approved versions of which are given below.  One was a revised set of
CLAS rules for taking shifts.  This set of rules was shaped with the
input of all the CLAS COPs, now known as PDLs.  It was approved
unanimously by CLAS members in the room at the time of the vote. The
other was a new bylaw to regulate the process of turning an approved
physics result into a publication which has the blessing of the whole
collaboration.  The draft was developed by R. Schumacher, B. Berman,
R. Minehart, and B. Raue.  At the end of the long discussion and some
revisions in the wording, the bylaw was passed unanimously by the CLAS
members in the room: the vote was 21 to 0.


------------------------------------------------------------------- 

 
                     CLAS DATA-TAKING SHIFT POLICY
                             Bylaw A1 

(As amended and approved by unanimous vote at the CLAS Collaboration
                         meeting of 1-22-2000.)
 
I. General Policies
-------------------
CLAS shifts are manned by two people at all times, one shift expert
(or "leader") and one shift worker. Expert shifts start at 00:00,
08:00 and 16:00 and worker shifts start at 04:00, 12:00 and 20:00.

Each collaborating institution has a contact person responsible for
filling all shifts assigned to that institution. The method for
choosing shift personnel is left to the discretion of the contact
person and his or her institution, but must conform to the overall
guidelines described here.  The current shift schedule and
shift-schedule editor are available at

      http://claspc1.cebaf.gov/SHIFTS/

This document only deals with general shift-taking concerned with
experimental data-taking.  It does not deal with specialized
shift-taking concerned with a particular detector or software element,
nor with any other activity performed by collaboration members in
fulfillment of their MOU's or physics program.
 

II. Shift Taker's Qualifications and Conduct of Operations
----------------------------------------------------------

All "Expert" and "Worker" shift takers must meet certain criteria
itemized here.  In addition, anyone participating in data taking with
CLAS must have a level of safety and radiation training described in
the CLAS "Conduct of Operations" (COO) document.

The Conduct of Operations (COO) document provides an authoritative
discussion of shift operations, including the duties and
responsibilities of the shift Expert and the shift Worker.  The COO is
presently available at

      http://www.jlab.org/Hall-B/ under "Current Run Information"

	Workers
	-------
All full and term members of the CLAS Collaboration are eligible for
worker shifts. Limited members and non-members, including new graduate
students, may be assigned shifts by the institutional contact persons
supervising the filling of the shift schedule.  Undergraduate students
may not be assigned shifts without prior approval of the Physics
Division Liaison (PDL, a.k.a. CLAS COP).

Prior to taking his or her first-ever scheduled shift, a person must
be an observer on at least one shift in order to become familiar with
procedures and (typically) to finish reading the required documents
listed in the COO.

	Experts 
	------- 
The list of expert shift takers who can act as shift leaders is
determined by the PDL, who is responsible for the safe and effective
operation of the system.  Experts must be full or term members of
CLAS.  Expert status implies experience in running CLAS shifts, not
necessarily expert knowledge of hardware and software components of
the system.

Oral communication skills that enable effective communication with the
accelerator staff are required for shift leaders.  The PDL must be
satisfied that any given shift leading expert has adequate language
skills.

If an individual is not qualified as an expert, the shift schedule
page will not accept him or her for an expert shift.


III. Shift Scheduling
---------------------

At six month intervals, the collaboration Shift Scheduler will revise
the membership list of the collaboration maintained by the Membership
Committee.  Shifts will be assigned in 4 day blocks to all member
institutions in proportion to their number of full and term members.
Limited members will NOT be included in this enumeration.

A shift-trading period will be organized under the leadership of the
Shift Scheduler to allow foreign, domestic, and local institutions to
optimize their periods of shift-taking responsibility to their needs.

Any persons added to an institution's list of Worker shift takers by
the institution's contact person will NOT thereby automatically become
members of the CLAS collaboration.  For example, new graduate students
who are assigned shifts are not thereby made members of CLAS.  By
assigning shifts to non-members, an institution's contact person is
certifying that the persons so assigned will have met the Worker shift
taker's requirements by the time the shift starts.

The PDL has the authority to reject a person for a shift if he or she
thinks that the person is unqualified.  The relevant institution then
has responsibility for finding a suitable replacement.

If a shift goes unfilled, either by no-show or no-qualify, the
incident will be brought to the attention of the Membership Committee
for possible action, such as removing institutional shift credits.

-----------------------------------------------------------------------
END OF CLAS SHIFT POLICY
-----------------------------------------------------------------------

_________________________________________________________________________

               Regulation of the CLAS Publication Process.
                             Bylaw C3 
(As amended and approved by unanimous vote at the CLAS Collaboration
                         meeting of 1-22-2000.)

Introduction: Section XI of the CLAS Charter provides the general
rules for publishing results in peer-reviewed journals.  The language
in that section was adopted as a charter amendment in January 1998.
This bylaw is intended to fill in the procedural details of this
process.

1) When is a scientific result ready for publication by CLAS?

The Physics Working Groups (PWG) have the initial responsibility to
recommend to the Coordinating Committee (CC) when a new CLAS result is
ready to be published.  The definition of, and the procedures for
releasing new scientific results itemized in Bylaw D2 (on Talks) cover
this first stage of the publication process.


2) When should an Ad-Hoc committee be designated, and how long should
it have to do its work?

The CC will appoint an Ad-Hoc committee as soon as it is informed that
a paper is being written.  It is recommended that lead authors ask for
an Ad-Hoc committee for their paper as soon as practical, in order
that the concerns and criticisms of the committee can be addressed in
a timely way.  The committee can work with the authors to discuss
issues related to publishability of a paper.  When the Ad-Hoc
committee is asked to review a finished draft of a paper, it should
act within 15 days with recommendations regarding the paper.


3) What should be the composition of the Ad-Hoc publication
committees?

The charter calls for an Ad-Hoc committee of usually 3 individuals to
be designated by the Coordinating Committee (CC) to review each paper
presented to the CC for consideration.  The Ad-Hoc committee shall be
composed of persons not directly involved in the results to be
published, and will have representation from at least two of the CLAS
Physics Working Groups.  At least one member shall be from the working
group which is developing the paper.  The members of the Ad-Hoc
committee shall be announced to the Collaboration.


4) The paper will be available for review by the Collaboration in a
secure (password-protected) area of the CLAS Web pages.  No member of
the Collaboration may disseminate the draft paper to persons outside
the Collaboration until final Collaboration approval for publication
is granted. In particular, distribution to e-print servers is not
allowed until collaboration approval is granted.

5) The comment period for feedback from the Collaboration to the
authors and Ad-Hoc committee shall be at least 15 days.

6) When is a paper approved for publication?

If there are NO comments from the Collaboration during the specified
comment period, approval is granted, according to the charter.  If
there are one or more criticisms, then the authors must address these
criticisms.  The Ad-Hoc committee will decide whether to accept the
responses, and whether to restart the comment period clock.  In case
of conflicts, the appeals mechanism in charter section XI.6 will be
followed.

7) Authorship.

In general, it is expected that authors on papers made some specific
contribution to the results described beyond taking shifts.  By
default, all full and term members at the time the data were obtained
will be included in the author list.  Members who wish to remove their
names should inform the lead authors before the end of the comment
period.  Former full and term members of CLAS will be included as
authors, if they wish, if they were members during the relevant data
taking period.  Members who joined after a particular data set was
obtained are included if they have made identifiable contributions to
the analysis of the data.

The list of authors is to be alphabetical from A to Z.  The lead
authors, however, will be listed first, in an order of their
choosing. The list of lead authors will be reviewed by the relevant
PWG.  In cases where results of a Ph.D. project are published, the
name of the student is to be listed first.  Limited members added at
the discretion of Spokespersons will be listed in the main body of the
list.  Disputes are to be mediated by the Coordinating Committee.

----------------------------------------------------------------------
End of Bylaw on publications.
----------------------------------------------------------------------
 
______________________________________________________________________


Any changes, additions or other clarifications of these minutes by the
Collaboration members will be cheerfully considered.  Amended minutes
will be sent out only in case of egregious errors or omissions.

[...]=chairman's editorials