All servers comply with version 2.0 of the Z39.50 protocol, though only 3 of the eighteen servers satisfied the requirements for supporting version 3.0. MARC format is the default syntax of choice, which all but one of the 18 vendors supported. None of the other syntax formats, including OPAC, SUTRS and GRS-1, was supported by more than four vendors.
All vendors support base Z39.50 services, such init, search and present. They also support Boolean AND and OR. Three of the 18 vendors do not support Boolean NOT, however. From here, more advance services drop off. For example, ten of 18 vendors have the Scan facility and many of these provide it for only a restricted set of indexes. Only two servers could sort result sets.
The preferred message length of records ranged from 30,000k to 5,000,000k. Maximum message lengths ranged from 30,000k to 10,000,000k.
In our attempt to categorize the ability to search Z39.50 servers, we have arranged indexes into 2 groups: 1st level searching, which meet the basic search needs of most library patrons, and 2nd level searching, which meet some of the expanded searching needs of librarians and catalogers.
1st level searching results:
Of the servers tested, one out of three servers performed searches on all the use attributes shown above in the table. Half of the servers could perform searches on most of the attributes and three servers supported only a few of the attributes. In some cases, a server would present a result set for a certain attribute, even though the attribute was not indexed. For example, some servers point all title-related queries to the title index. Queries on Title Uniform (5) and Title Series (6) would actually search the title index (4).
Also, not shown in the table, are the structure attributes supported by the servers. The structure attribute indicates the different ways a field in a record is indexed. The most common structure attributes supported by the servers are keyword (1) and phrase (2). However, not all servers supported these structure attributes across all available indexes.
Right-hand truncation, truncation attribute (1), is supported by roughly half of the servers. Few servers support this attribute across all available indexes.
Almost half of the servers supported at least a few of the second level searching use attributes, three servers supported searching of most of the indexes and one server supported all of the indexes. As with first level searching, some servers link different second level attributes to the same index. The most common use attributes supported in these tests include ISBN (7), ISSN (8), Date (31), Corporate Name (2) and Conference Name (3).
Some servers direct searches of unsupported use attributes to a default index. Most of the servers do not support right-hand truncation or scanning of second level indexes. Some servers allow custom configuration of available indexes, though results may vary for different users.
The second level use attributes are just a subset of a much larger use attribute set. We restricted testing for the attributes defined above, although the servers may support other attributes. This is simply an indicator of Z39.50 development in the industry and what to expect from existing servers.
www.amlibs.com
CARL
Corporation
www.carl.org
Dialog
Corporation
www.dialog.com
DRA
-- Classic, Inlex, MultiLIS
www.dra.com/resources/z3950/draz3950.html
EBSCO
www.ebsco.com/home
Endeavor
www.endinfosys.com
Ex
Libris
www.exlibris-usa.com
Index
Data
www.indexdata.dk/zebra
Innovative
Interfaces, Inc.
www.iii.com
Ovid Technologies www.ovid.com/support/ztry/ztry1.htm
SilverPlatter
ERL
www.silverplatter.com
SIRS
Mandarin, Inc.
www.sirs.com