From: "Les Wibberley - CAS - ext. 2330" ZIG Meeting Gainesville Hilton Gainesville, Florida January 26-28 1994 Day 1 1) Introductions and Status Reports FCLA - Has production server providing access to databases for 9 Florida universities. Client is nearing production status. Working on attribute mapping issues. Terry Sullivan's Thinosi (TOSI) software has been put in the public domain. TOSI includes ACSE, Presentation, and Session layer functionality. University of Missouri -Issued an RFP for a client, received 7 responses, but procurement delayed til fall 94, due to funding. Meanwhile, working with NLC code. Penn State - Producton client is accessing RLG databases via RLG's server. Also production server. RLG - Working with Penn State, providing access to production server (35,000 sessions). Also supporting Eureka mainframe client (access via telnet) AT&T - Both internal and external servers in operation. DRA - Both client & server products released to customer sites. Client & server are separately licensed. Some licences have been sold. Windows client will be ready for testing soon. Test server available for testing, provides logging to support debugging. OCLC - Has production server providing access to 30+ DB's. SCAN now available on their development server. Also adding Z39.50 to PRISM. UC-DLA - Server is still a protoype, but supports 300-400 sessions per day. Lots of client work underway, moving server toward production. Using Z39.50 client for accessing DRA Citadel and FirstSearch. Has added support for OPAC Record Syntax. Will interoperate with RLG. Library of Congress - Mainframe client and server going into production around February 1994. Using OCLC SiteSearch system on RS-6000 to support Russian and German reference records. CMU - System has previously been based on Z39.50-1988. Have decided to convert to V2. Have a V2 Motif client interoperating with several servers, including AT&T, BSD, Penn State, Geac, LC, DRA, CNIDR... Next step is to convert the server to V2, then develop an image server using Z39.50. Dartmouth - Client & server system to go to beta/production in about 2 months. Gaylord - V2 client & servers are in the field, have sold & installed some licenses. MS/DOS client is in beta testing. NOTIS - Mac client is in testing. WAIS Inc - Developed a WAIS profile for using Z39.50 Version 2, which has been approved by the OIW SIGLA. Currently implementing & testing V2 using WAIS profile. Silver Platter - working with NOTIS, developing Z39.50 gateway to databases, mapping output to MARC records. GEAC - V2 target working. GUI-based Origin working. VTLS - client released, working on server. Apple - Apple Search is moving toward support of a WAIS Gateway. CNIDR - working on MS Windows client project, using DLL protocol library. Have a WWW-Z39.50 gateway to their server. Mead Data - have a couple of current awareness products in Beta; accessed from OS/2 Origin; profile public. Mapping a natural language interface to Z39.50 via RPN Query. Chemical Abstracts Service - continuing development of Z39.50 server. Participating in Z39.50 pilot project with US Patent Office; CAS server interoperating with the CNIDR client. Adding support for several V3 features to server. ESL - has a couple of client projects underway. U. of Wisconsin - using NLC code to build a gateway to their catalogs. Going to Beta in February. Dialog - currently working on a limited V3 Z39.50 client & server. Will be publicly available in 1995. Interested in Extended Services. National Library of Canada - have put their Z39.50 client/server software in the public domain. Several other organizations are using the software, and NLC is receiving lots of good feedback. Software Kinetics - building a new Z39.50 bibiographic system for NLC. Conquest Software - developing a profile for using their natural language search system for a Z39.50-based government Thesaurus project. Name Organization Email Address ~~~~ ~~~~~~~~~~~~ ~~~~~~~~~~~~~ Eric Bivona Dartmouth College eric.bivona@dartmouth.edu George Blair Conquest Software conquest@access.digex.net Dave Brooks Idaho Nat'l Engineering Lab dab@inel.gov Bette Brunelle CD Plus betteb@cdplus.com Paul Buchanan Washington University buchanan@library.wustl.edu Chris Buckley Cornell University chrisb@cs.cornell.edu Don Burns Auto-Graphics dwb@agaviion.agfx.com Bill Campbell Dialog campbell@dnt.dialog.com Scott Daup Westlaw daup@research.westlaw.com Ray Denenberg LC ray@rden.loc.gov Sean Donelan Data Research sean@dra.com Parviz Dousti CMU dousti@andrew.cmu.edu Curt Ellmann University of Wisconsin curt@adp.wisc.edu Eric Ferrin Penn State egf@psulias.psu.edu Rich Fuchs RLG rbf@lyra.stanford.edu Kevin Gamiel CNIDR kevin.gamiel@cnidr.org Mike Gursky CD Plus mgursky@cdplus.com Roger Hall Dialog hall@dnt.dialog.com Mark Hinnebusch FCLA fclmth@nervm.nerdc.ufl.edu Craig Jackson JAC craigj@epub.ziff.com Kurt Kopp University of Missouri koppk@ext.missouri.edu John Kunze UC-Berkeley jak@violet.berkeley.com Ralph LeVan OCLC rrl@oclc.org David Loy Dialog loy@dnt.dialog.com Cliff Lynch University of California clifford.lynch@ucop.edu Denis Lynch ESL Inc. dml@esl.com Jim McDonald OCLC jhm@oclc.org Brad McLean Gaylord Information Systems brad@saturn.gaylord.com Jim Michael DRA jim@dranet.dra.com Bill Moen Syracuse University wemoen@rodan.acs.syr.edu Mark Needleman UC-DLA mhn@stubbs.ucop.edu Andy Oates Geac Computers a.oates@geac.com Ralph Orlik LC orlik@mail.loc.gov Cecilia Preston cpreston@info.berkeley.edu Sara Randall NOTIS srandall@notis.com Peter Ryall MDC peter@meaddata.com Mary Schonsberg Idaho Nat'l Engineering Lab srf@inel.gov Andy Shapiro Silver Platter andys@silverplatter.com Philip Stecha VIS Consultants vis@access.digex.net Lennie Stovel RLG bl.mds@rlg.stanford.edu Margaret St. Pierre WAIS Inc. saint@wais.com Terry Sullivan FCLA fclts@nervm.nerdc.ufl.edu Margery Tibbetts UC DLA mztuc@uccmvsa.bitnet Fay Turner National Library of Canada fay.turner@ncl-bnc.ca Howard Turtle Westlaw turtle@research.westlaw.com Janet Vratny Apple Library janet@apple.com Bob Waldstein AT&T Library Org. wald@library.att.com Les Wibberley Chemical Abstracts Service lhw24@cas.org Cathy Winfrey VTLS, Inc. winfreyc@vtls.com Joe Zeeman Software Kinetics zeeman@sofkin.ca 2. Future meetings ~~~~~~~~~~~~~~~ OCLC offered to host a 2-day Z39.50 tutorial in March/April 1994 timeframe. The ZIG had originally planned a 2-day tutorial prior to April ZIG meeting. But nothing had been organized yet. Sara Randall agreed to help organize a tutorial if needed. Ralph LeVan said that OCLC needs to train several of their staff anyway, and offered to coordinate the tutorial with ZIG, charging a small fee per attendee to cover overheads. Members of the ZIG would be invited to be presenters, and would be compensated for their expenses. OCLC could provide capacity for about 100 attendees. The goal of the tutorial would be to provide training for technical folks to implement Z39.50. After much discussion, the following schedule for meetings was agreed to: o March/April 1994 - Z39.50 tutorial at OCLC (Ohio) o April 27-29 1994 - ZIG meeting hosted by NIH/NLM (Wash. DC) o September 19-20 1994 - Z39.50 tutorial at CNIDR (N.C.) o September 21-23 1994 - ZIG meeting at CNIDR (N.C.) o January 11-13 1995 - ZIG meeting at RLG (CA) In addition, the ZIG will consider sponsoring a tutorial in conjunction with the January ZIG in California, to provide easier access from Pacific Rim attendees. 3. V3 Balloting Plans ~~~~~~~~~~~~~~~~~~~ Ray Denenberg, Z39.50 Tech. editor, stated following V3 balloting goals: o create stable, ballotable draft by mid-1994 o get into balloting process by June 1994 o 3-month balloting period o discuss ballot objection resolution at 9/94 ZIG meeting o complete balloting & obtain approval before end of 1994 Ray noted that V3 might go to ballot before it is fully complete (there may be some loose ends to clean up, such as details of diagnostics). Ray particularly requested any input on additional diagnositics as soon as possible. Discussion of V4 topics is already starting. Ray thought that development and balloting of V4 could proceed more quickly than V3 has. 4. Register of Implementors ~~~~~~~~~~~~~~~~~~~~~~~~ Ray distributed an updated Register of Implementors (ZIG94-102). There are now about 53 registered implementors of Z39.50. Attendees were asked to mark any updates to Ray's copy. Ray may drop some obsolete entries, unless he hears from the organizations. New organizations are encouraged to register if they are actually implementing Z39.50. Ray agreed to make softcopy of the Register of Implementors available in Postscript for ftp access. 5. Z39.50 V3D9 ~~~~~~~~~~~ Ray distributed the latest V3D9 draft of Z39.50 (ZIG94-101). Ray noted that V3D9 is the last numbered draft in the V3 series. V3D9+ will contain some minor corrections and a more detailed writeup of the conformance statement. The next draft will be called the V3 PBD (Pre-Ballot Draft). Ray noted that Z39.50 drafts will continue to be provided in Postscript via anonymous ftp and in ascii from LC Marvel. Ray also noted that the ASN.1 for Z39.50 V2 is now available from LC Marvel. 6. NISO report ~~~~~~~~~~~ Cliff Lynch has joined the NISO board, due to his interest in the issue of electronic distribution of standards. The NISO board has agreed that both drafts and final forms of their standards should be available electronically. They established a task force to build an implementation plan for this, and were to meet on 1/28/94. Cliff hopes that electronic distribution of NISO standards will be supported by the end of 1994. This support will start with some of the more current active NISO standards, such as Z39.50, then look at supporting older, less active standards. Initially, ASCII versions will be supported, followed by postscript. FCLA offered to scan in older NISO standards, if they weren't available otherwise. The standards will carry a NISO copyright statement, to control use and distribution. It was noted that it will be important to insure that there is no restriction on extracting the ASN.1 from the softcopy standard, and using it to build products. Cliff may post a message to the ZIGlist documenting outcome of the 1/28 meeting. 7. V3D9 Walkthrough ~~~~~~~~~~~~~~~~ Ray reviewed the changes incorporated in V3D9 since V3D8. This walkthrough triggered discussion on a wide variety of topics. A summary of the key points discussed is documented as Attachment 1 to the minutes (not here). 8. GRS/ES/variants/schema ~~~~~~~~~~~~~~~~~~~~~~ Ray provided an overview presentation on the new material in V3D9, covering the concepts and details of the PresentRequest CompSpec facility, the GRS-1 Record Syntax, the ES-1 Element Specification, the Variant-1 specification and the Schema-1 definition. This presentation triggered discussion on a wide variety of topics, and lasted for most of Day 2. A summary of the key points discussed is documented as Attachment 2 to the minutes (not here). 9. WAIS Profile ~~~~~~~~~~~~ Margaret St. Pierre provided copies of the WAIS profile (ZIG94-105) for using Z39.50 Version 2, which has been approved by the OIW SIGLA. Discussion ensued, intertwined with the topics from Ray's presentation, summarized by the following. The meaning of rank in this profile is weight. It was noted that both rank and weight are needed. Ray will update Schema-1 definitions to reflect both rank (relative position in result set) and score (weight). For a set of N records, rank would range from 1-N. While score/weight is a record's relative relevance. ESN is case-insensitive. WAIS will change ESN 'V' to 'WAIS-V' to avoid namespace collisions. Same with 'G' for GILS profile. Note: ESN cannot be structured in standard - but can be constructed by specific WAIS (or other) server.. John Kunze suggested that commonly used elements be defined in Schema-1. For example, the elements defined in the Summary Record. This idea is consistent with John's previous proposal for the Info-1 generic attribute/element set. Les suggested not putting all common elements in Schema-1; instead, define a Schema-2 (avoids overlaps in schemas), and allows an implementation to import Schema-1, separate from common elements in Schema-2. John will write up a proposal for new schema including common attrs/elements (meta & real). 10. Attributes ~~~~~~~~~~ There were a large number of specific issues discussed under this topic. A summary of key points follows. Cliff noted that implementation experience to date indicates that the single largest Z39.50 interoperability problem is related to interpretation of the bib-1 attribute set. This must be addressed. One problem is that there is inadequate documentation on the meaning of the attributes which have been defined. There is currently only a separate non-normative out-of-date document which helps provide some explanation for interpreting attributes. There are two specific areas which need work: (1) definitions of USE Attributes and their mappings to MARC; and (2) the relationships between attributes, their semantics, rules for use, and defaults. Cliff asked: do we want to expand descriptions in Appendix ATR (the names are very cryptic). Ray: Would have to be non-normative - maybe write a separate tech report not included in std (std getting too large). Cliff suggested: put expanded info in separate doc but basic desc. in ATR. Ray said: best solution is to put in Implementor's Guide, including attr specifics, combination rules (this summer, after V3 goes to ballot). Mark suggested that some generic rules for attribute set behavior (multiple use attrs, etc.) should go in the V3 standard. There are several key parts to defining normative behavior: 1) multiple occurrences of an attribute type (use/struct/etc) 2) attribute defaults (e.g, if attr omitted), and how specified 3) completeness, position - what do they mean in general? 4) word list (semantics, interpretation - loose, tight?) 5) the meanings of attributes and how to interpret Others suggested that the standard remain silent on these issues. One basic problem is that specific attribute semantics cannot now be easily retroactively defined, when they were previously undefined, since that would break many implementations. Need to document what happens when an attribute is omitted (any type). It was agreed that if a client omits a given attribute type, then it is the server's choice to select the default behaviour. Should a server ignore an attribute which it doesn't support, or reject the search? No agreement reached on this, since implementations and needs vary greatly. It was suggested that the client needed to be able to convey information to the server about how to handle attribute interpretation (for example, ignore or reject unsupported attributes). This might be accomplished by defining specific fields to be carried within the AddSearchInfo in the SearchRequest. A proposal for an AddSearchInfo structure and its fields will be written up. This may also be useful in specifying the appropriate balance between natural language vs. strict boolean query interpretation. It was also generally agreed that this client direction for server interpretation is needed at both the detailed (attribute) and the global (query) level. Peter brought up fact that he had proposed a new attr type a year ago & never received any comments (+/-) & it was not accepted with no reason given. This is example of how hard it is to add new Attr type. Ray suggested that Peter re-propose. Peter's new attribute type: morphology attribute type to include values such as: caps, allCaps, noCaps, singular, plural, soundex? (possibly move phonetic & stemming over?). There were many additional attribute details discussed, which Ray will address in the next draft. - end of day 2