CERN Accelerating science

This website is no longer maintained. Its content may be obsolete. Please visit for current CERN information.

CERN home pageCERN home pageDocuments by ReferenceDocuments by ReferenceCNLsCNLsYear 2001Year 2001Help, Info about this page


Editorial Information
If you need help
Announcements Physics Computing Desktop Computing Internet Services and Network Scientific Applications and Software Engineering Desktop Publishing The Learning Zone User Documentation Just For Fun ...
Previous:The Design and Implementation of Large Packages: Hbook, GEANT3, PAW, ROOT
Next:EUCLID: the Beginnings of Computer Aided Design
 (See printing version)

The PAW History Seen by the CNLs

Olivier Couet , IT/API


This article shows how the "Physics Analysis Workstation" (PAW) program evolution has been seen accross CNL editions.

From the initial announcement of PAW in 1986, until the Y2K modifications done in its code thirteen years later, PAW has appeared in almost 30 Computer News Letters. We thought it would be interesting to quickly scan all of the articles and obtain an overview of the "PAW History as Seen by the CNL".

PAW was first mentioned in CNL 186 at the end of 1986. "A schematic comparison of HTV and its successor PAW" was presented at this time. PAW was in the testing phase and had not yet really been put to use. Certain of the PAW basic components, HBOOK and HPLOT, were already present in HTV. In this schematic view, present were almost all of the fundamental PAW elements still used today: HBOOK (histogramming), HPLOT (plotting package), HIGZ (basic graphics), ZEBRA (memory management and I/O), KU (user interface). Although, for reasons unexplained in CNL 190 (1987), KU would then be renamed KUIP. A few months after the initial announcement, the first test release of PAW was publicized in CNL 188 in May 1987 and was only for "selected test users". The test period lasted just a few months, and in mid 1987, CNL 189 announced the first (non-test) release of HBOOK 4.00, HPLOT 5.00 and of a new package called HIGZ 1.00 (now at release 1.27/02). Based on these recently released packages, the first official PAW release on Apollo, VAX and IBM VM/CMS was announced in CNL 190 at the end of 1987. This was also the first release of KUIP. Compared to the schematic presented in CNL 186, two packages were added: the Fortran interpreter - COMIS and the arrays manipulation package - SIGMA. The minimization package MINUIT was not mentioned at this time.

In CNL 190 also, a very simple "Hello World" style PAW example was presented. It was suggested to users to create a PAWLOGON.KUMAC file with:

MESSAGE 'Hello, this is my logon macro executing'.  
And the article said: "Once your logon macro is executed you get the prompt:
Amazing isn't it ? ...

An article of 1987 began with the sentences: "PAW is conceived as an instrument to assist physicists in the analysis and presentation of their data. It provides interactive graphical presentation and statistical or mathematical analysis [...]". This description of PAW has since been used many times in various articles and PAW presentations, and still today is the first piece of text one can find on the PAW home page. The article also explained how to install PAW and its related packages from PATCHY PAM files.

The next summer (1988 - CNL 192), a PAW meeting was announced to "obtain first experience from users and suggestions". The PAW version was 1.3 (still on the 3 previous platforms). It had an improved version (the first) of the "Ntuple Selection Mechanism and cuts". In the next CNL (194 January - March 1989), an HBOOK example was given showing how to create an Ntuple in a batch program and then perform an analysis on it utilizing PAW. This Ntuple is now part of the standard PAW examples.

Before mid 1989 (CNL 195), PAW was distributed only as an executable. Since then, the PAWLIB library is available containing PAW callable routines, COMIS and SIGMA. This allows access of these functionalities for any user programs or to build special PAW versions (more memory, experiment specific commands, etc.). The discussion list LPAW is created only at the end of 1990 (CNL 200). Later, the HEPLIB news group would replace this list.

With the user community of PAW growing, it then went into a production phase. At the beginning of 1991, the version number was 1.11. The X11 graphics driver (even on VM) had more and more importance even though a "New Falco driver" was announced in CNL 201. At that time, graphical user interfaces became more and more common. Motif was the main system in this area, and the first version of PAW using Motif appeared in CNL 203 (late 1991). It was not yet Paw++ but that first attempt was a good prototype to understand what could be done and what needed to be accomplished. It was in the same CNL that an improved version of KUIP allowing it to perform DO loops, IF-THEN-ELSE etc. in the macro scripting language was announced. At the end of 1991, the "World Wide Web" appeared for the first time in CNL 204. The PAW web site will appear a couple of years later. Several improvements in HIGZ had been accomplished in summer 1992 (CNL 207). The number of graphics drivers was growing including Mac and MSDOS.

CNL 209 in late 1992 announced big modifications in PAW that would later be explained in CNL 210 in the longest article ever published on PAW in the CNL. These news items included: Paw++, the Column Wise Ntuple and Kuip/Motif. The new functionalities were in such an amount as to justify an increment of the main digit of the PAW version number, hence the PAW version number is now 2.00! In that CNL, PIAF's concepts were also explained. The first PIAF cluster would be in production almost a year later on five HP 755 (CNL 213) together with a modernized version of the Ntuple query processor (based on LEX and YACC).

Between these major changes some more minor events occurred in the "PAW world". From CNL 211 to 217 (March 1993 - July 1994), a test kit was set up, PAW was ported to Linux, the current HEPLIB discussion list started, KUIP rewritten fully in C, and also during that period, CERN reached its 40th anniversary.

Appearance of PAW on the WWW was communicated in CNL 218 at the end of 1994. That first version of the PAW web site contained items still existing on the site today. In particular, a list of seven FAQs was already available (we have 185 today). These same questions are still "frequently asked" because they are still there in modified and more complete forms. A PAW questionnaire was also launched in that CNL, attempting to better understand how PAW was being used, and the first attempt to manage PAW code with CVS was mentioned in this article.

Even if the second version of the Ntuple Query processor was much improved over the first, it was not powerful enough to handle the power of the Column Wise Ntuples. A new version was requested by users and appeared at the beginning of 1995. With the new Ntuple query processor, the last weak element of the system disappeared and PAW entered a consolidation phase. This was also visible on the PAW web site where the new sections "PAW usage", "Tutorial" and "Reference manual" appeared during mid-1995 and the users section "Contributions" at the end of the same year (CNL 220). The "end of PAW" was announced in CNL 223 (June 1996). It was the "last major PAW release" and the people involved in the PAW development and support were directed to concentrate on "the gathering of user requirements for a future data analysis tool, including requirements for the components, e.g., histogramming, minimization, data input", a sad occasion for the PAW developers.

Since CNL 225 (December 1996), PAW announcements can be found only on the WWW. The last reference of PAW in the CNL was in CNL 237 (December 1999) telling users that they should not worry about a Y2K bug affecting PAW, it should go over the new year (millennium) border safely... and it did. So, since 1999, PAW activities are no longer reflected in the CNL. However, all information can be found on the PAW web site. Upon visiting there, you may find that it was not the "end of PAW" as CNL 223 predicted in 1996.

For matters related to this article please contact the author.

Vol. XXXVI, issue no 1

Last Updated on Thu Apr 05 15:28:11 CEST 2001.
Copyright © CERN 2001 -- European Organization for Nuclear Research