Z39.50 Draft Attribute Architecture

February 18, 1998

An earlier version of this draft attribute architecture was discussed at the January 1998 ZIG meeting. Notes from that meeting were applied to the February 6 version of this document. Comments were solicited until February 17. Those comments have been applied to this (February 18) version.
This document will remain stable until after the NISO attribute meeting in March. The document may subsequently be revised as a result of that meeting.

Following are changes applied between February 6 and February 18.

At the beginning of 3.2.1.1,
Change:
When multiple Use-type attributes occur within an operand, nesting is implied, unless a semantic indicator is present, indicating a different interpretation of the multiple occurrence.
To:
Multiple Use-type attributes may occur within an operand, and this may imply nesting (either implicitly or via semantic indicator; multiple occurrences may have a different interpretation, also either implicitly or via a semantic indicator).
Section 3.1.5.1 has been added.
The latter portion of 3.2.1.2 has been re-written to lift the restriction for a nested search that the path is assumed to be anchored.
Added a reference in 3.2.3 to "character set and language negotiation".
Globally changed "character-valued" to "string-valued".

1. Introduction and Preliminary Notes

1.1 Historical Background

Ed. Note: Historical background to be supplied.

1.2 Acknowledgements

Ed. Note: List of meeting participants and other significant contributors to be listed here.

1.3 Brief Technical Background

Z39.50 specifies a number of query types, and requires support for one of those types, the type-1 query. This document is concerned with the Z39.50 type-1 query only.

The type-1 query consists of one or more search terms, each with a set of attributes, specifying, for example, the type of term (author, title, subject, etc.), whether the term is truncated, and its structure. The server is responsible for mapping attributes to the logical design of the database.

A term in a type-1 query, together with its accompanying collection of attributes, is referred to as an operand. Operands may be combined in a type-1 query, linked by boolean operators (And, Or, And-not, and Proximity).

Each attribute is a pair: an attribute type and a value of that type. An Attribute set defines a set of attribute types, and for each, a list of possible values.

Example: The bib-1 attribute set defines a number of attribute types; one of which is Use. For bib- 1 Use attributes, many attribute values are defined, one of which is personal name.

An attribute set definition is assigned an object identifier, referred to as its attribute set identifier.

In version 2 of Z39.50, all attributes within a query must belong to a single attribute set. In version 3, attributes may be combined from different attribute sets, within a single query, even within a single operand. This is a significant enhancement, providing support for multiple database searching, and allowing attribute sets to be defined with less replication.

Also in version 3, new data types for terms are defined (in version 2 only binary values are allowed).

1.4 Version 3 Assumption

There are several enhancements in version 3 pertaining to attribute sets and query construction; the two enhancements described at the end of 1.3 are clearly the most important, and are seen to be functional prerequisites for attribute architecture. For this reason, version 3 is assumed by this architecture, and version 2 is not addressed.

1.5 Limitations

The Z39.50 type-1 query has known limitations, and the architecture specified in this document is restricted by these limitations. As the standard evolves and new versions are approved, the architecture may be expanded.

1.5.1 Semantic Indicator

In order to compensate for some of the type-1 limitations, it may be necessary to utilize the semantic indicator (provided within version 3) for purposes that would otherwise be accomplished by more coherent mechanisms if these limitations were not present. It should be thus noted that in future versions of Z39.50 it is intended that these limitations will be addressed, obviating the need for extensive use of the semantic indicator at the attribute level.

1.5.2 Nesting and Occurrence

When specifying, for example, "field-1 within field-2" it will be possible to specify an occurrence, for example, "second occurrence of field-1, within field-2" but not, for example, "second occurrence of field-1, within second occurrence of field-2". That is, only a single node may carry occurrence, likely either the root or the leaf, and this should be statically specified within the attribute set definition. (This is a limitation of class 1, not a limitation of the type-1 query, and may be overcome by classes other than 1).

1.6 Some Unresolved Issues

Though there is clear consensus within the ZIG that nesting of fieldname attributes should be supported, there is division of opinion on whether nesting should be permitted for Use attributes.
Similarly, there is division of opinion on whether specification of occurrence should be permitted for Use attributes.
some feel that the limitation described in 1.5.2, for occurrence, can be overcome, for example by use of the 'complex' form of attributeValue.
It is not clear whether anchoring is sufficiently specified. See 3.2.1.2.
3.1.2 states that a class 1 attribute set may not define any attribute types not defined for class 1 (i.e. not defined in this document). Some feel this is overly restrictive.

2. Attribute set Class definitions

An Attribute set class definition provides an umbrella context for the definition of an attribute set belonging to a particular class. A class definition defines all of the attribute types that may be included in an attribute set for that class.

(At least one attribute set class definition will be developed, but it is not clear that more than one will be necessary.)

3. Attribute Set Class 1

This class is intended to cover all known, existing needs (existing attribute sets will need to be re-specified within this framework). The intent is not to preclude new types of attributes beyond those specified here; it should be possible to add new attribute types to this broad attribute class, if they are relatively orthogonal to the attribute types defined here.

Note: There may be other approaches developed which partition the set of attributes into fundamentally different types; this might result in the definition of a new attribute class that is inconsistent with this class. However, no need for such a separate class has been identified.

The importance of enumerating all of the possible attribute types within this "universal" attribute class is to provide a template for developers of attribute sets, and to set up a framework for interoperability among independently defined attribute sets which are intended to serve various communities. In particular, it should be possible for groups of content experts to develop new use attribute sets, ASN.1 datatypes, comparison operators, and perhaps structure/format attributes which fit comfortably within this framework. Server developers can, based on the template defined here, recognize various attribute types that are omitted in a given query, as well as illegal repetitions or combinations of attributes of given types that would indicate a malformed query.

3.1 General Rules for Class 1

3.1.1 Semantic Precedence and Interaction among Sets

The context of the attribute class defined here would be identified as being in effect for a query by specifying the OID of an attribute set conformant with the class in the global OID for a Z39.50 query (most likely one of the utility attribute sets which it is proposed below that the ZIG develop). The "global" OID refers to the object identifier within the type-1 query that does not accompany a specific attribute. For class 1, this is referred to as the dominant OID for the query. When attributes from different attribute sets are mixed within a query, and when the respective attribute set definitions conflict such that the resulting semantics are ambiguous, the semantics of the dominant set prevail.

Any attribute set definition should:

indicate whether that attribute set may be used as the dominant set in a query; and if so:
describe the rules that apply, when that set is used as the dominant set, for intermixing of attributes from different sets within an operand or query.

Interaction between attribute sets conformant to this attribute set class and historical attribute sets not conformant to this class within a query operand are undefined.

3.1.2 Inheritance and Population

An attribute set consistent with this attribute class will define attributes of one or more of the types specified in 3.2.

Any class 1 attribute set inherits the rules prescribed for the class, that apply to attribute types defined for that set. However, a class 1 attribute set need not define nor populate every attribute type defined for class 1. A class 1 attribute set may define as few as one attribute type, or as many as all of the attribute types defined for class 1. It may not define attribute types not defined for class 1.

3.1.3 Omitted Attributes

If attributes of a given type are omitted in a query they should be treated as omitted in establishing the semantics of a given query (in other words, there are no defaults for omitted attributes).

3.1.4 Mutual Exclusivity

Some types of attributes (for example, use and field attributes) are mutually exclusive in a given query operand; these rules are defined at the level of the attribute class rather than specific attribute sets.

3.1.5 Repeatability

In general if any attribute is allowed to be repeatable, the semantics of repeating the attribute must be well defined (implicitly or explicitly).

When an attribute set definition is being developed and the need is foreseen for an attribute to repeat, for example when values are orthogonal, it is recommended that the developers consider separating the values into different attribute types, if possible.

While repeatability may be permissible for a given attribute type, as a general principle, an attribute type should not be repeated as a substitute for Boolean operations. To amplify this point, an attribute definition might prescribe how to interpret, for example, multiple Use attributes in a single operand. For example, the definition might prescribe:

Multiple Use attributes may be supplied in order of preference, so if a server does not support the first supplied, then use the second, etc.; or
if multiple Use attributes are supplied, the server is to choose the "best" among the set; or
multiple Use attributes implies nesting, thus if Use attributes use-1, use-2, and use-3 are specified in a single operand, in that order, it means search for use-3 within use-2 within use-1 (see "Nesting of Use-type Attributes" 3.2.1.1).

The definition may include a semantic indicator, allowing a client to select among several semantic alternatives. However, none of those alternatives should be to construct separate operands (linked by boolean 'and' or 'or') for each Use attributes. The reason is that the type-1 query supports boolean operations, so allowing another means of specifying boolean operations would add un-necessary complexity. This is in contrast to potential semantic interpretations of multiple Use attributes which cannot be otherwise represented via the type-1 query, as in the examples above.

3.1.5.1 Mechanism for Repeating Attributes

There are two mechanisms for providing multiple attributes of the same type within an operand:

Via 'list' within 'complex' CHOICE of 'attributeValue' within AttributeElement.
Via separate instances of AttributeElement.

The first mechanism (provided by version 3, and not supported in version 2) is the mechanism prescribed for this class.

3.2 Attribute Types defined within the Attribute Class

3.2.1 Use-type attributes

This attribute class definition recognizes that some applications of Z39.50 make a strong link to database schemes, while others continue to work with abstract definitions of databases. Thus there are two distinct attribute types to accommodate these very different approaches to the use of Z39.50. These two types should not be mixed within an operand.

Database Fieldname
Defined in conjunction with a specific database schema. It can be qualified via repetition of the attribute. An example of such qualification (nesting) might be a field path within an SGML database. The Fieldname attribute should not be used in conjunction with the Use attribute unless the attribute definition describes how they may be used in combination. This is generally a string-valued attribute, though it may also be useful to permit numeric values to facilitate mapping of fieldnames from schema definitions that use numeric assignments for fields.
Use attribute
Defines an intellectual access point for a group of relatively homogeneous databases, independent of database schema. Nesting, established through repetition, is also valid for the use attribute and establishes a context for the use attribute (for example, place names within abstracts). There is not a clean line between a qualified use attribute and the expansion/interpretation attributes discussed below. A good example of this is the creation context used by CIMI, which is an attribute on as SGML field. This should be a numerically valued attribute.

3.2.1.1 Nesting of Use-type Attributes

Multiple Use-type attributes may occur within an operand, and this may imply nesting (either implicitly or via semantic indicator; multiple occurrences may have a different interpretation, also either implicitly or via a semantic indicator).

The order of nesting should be as in the following example: if field-1, field-2, and field-3 are supplied, in that order, it means field-3 within field-2 within field-1. This rule, though arbitrary and perhaps beyond the scope of architecture, is supplied in order to avoid conflicting definitions, and reduce complexity of implementations supporting multiple attribute sets where nesting is prescribed.

3.2.1.2 Anchored vs. Non-anchored Searching

Whether a search is flat or nested (structured), it should be either implicit clear, or there should be an explicit indicator (see 3.2.8) designating whether the access point path is anchored or unanchored.

Definitions:

Anchored means that matching must occur from the root of the element tree;
unanchored means that matching may occur beginning at any node within the element tree.

example:
Suppose a schema has elements Name (unstructured) and Creator, structured into sub-elements Name, eMail, and Affiliation:

Name
Creator

Name
eMail
Affiliation

When fieldName attribute Name is specified, as anchored, then it is intended to match the first-level Name; if multiple-fieldname- attributes Creator and Name are specified, as anchored, then it is intended to match Name within Creator. If the single fieldname attribute Name is specified, as unanchored, then it is intended to match either Name or Name within Creator.

A single wildcard, for example, "Any", may be used in a fieldName path.

3.2.1.3 Mixing Fieldname Attributes from Multiple Attribute Sets

Mixing Fieldname attributes from multiple attribute sets is permissible, and no attribute set conforming to this class should preclude mixing of its fieldname attributes with fieldname attributes from other sets. This is a cross-attribute-set rule for any attribute set conforming to class 1.

3.2.2 Query Management Type Attributes

These attributes have the property that they can be rewritten by the server as part of the return of a revised query returned back to the client as additional search information.

Weight
The weight of an operand in a weighted boolean query. This should be registered (along with a normalized value range) as part of a basic attribute set within the attribute class. This is a non-repeating numeric attribute.
Hit Count
The number of records satisfying the operand. This should again be registered as part of the basic attribute set. This attribute is intended for purposes of conveying information from server to client, but it may passed back from client to server (when the client simply wants to turn around a reformulated search -- in that case, it is to be ignored by the server). This is a non-repeating numeric attribute.
Stopwording
For a query sent from client to server, may be used to request that the server not treat a given word as a stopword. For a query returned from the server, may be used to indicate that a given word was treated as a stopword.

3.2.3 Qualifying Attribute Types

Language Attribute
The language of a given query operand. A general-use version of the language attribute, using values defined from some standard source should be defined and registered so that it is generally available. It is not clear whether a string to numeric mapping is needed for this attribute type or whether it should simply be string-valued. Not repeatable.
Note: A Character set Attribute is not proposed; the current thinking is that it is unnecessary since it is handled by general Z39.50 character set support.See Character Set and Language Negotiation
Content Authority
The source of the term. This is a string-valued attribute. In the interests of simplicity, probably should be non-repeatable, although there may be situations where repeatable content authority could be meaningfully interpreted.
Expansion/interpretation
Indicates that thesaural expansion, singular/plural matching, part of speech qualification, phonetic matching, case sensitivity, or stemming should be used in the query evaluation. Word by word truncation is also viewed as a form of stemming and is to be included within this attribute type, as would various loose forms of phrase marching. Repeatable; may be string or numeric valued.

3.2.4 Comparison Attribute Type

There are different comparison attributes for each of the term-value datatypes discussed below. (See also the discussion of datatyping for operand values below.)

Comparison attributes are strongly typed. They are mandatory, non-repeatable and numeric valued.

Comparison attributes are somewhat similar to the relation attributes of bib-1, but named differently to avoid confusion. Note that equality is used only for cases of true equality testing (i.e. string and integers). Various "matching" comparison operators are used for string matching using various kinds of regular expressions, for example. Sample values might include:

complete match
doesn't match
left and right anchored match
contains
contained in bounding-polygon
match via grep
relevance feedback
equality as strings
numeric greater than
between (range operations in conjunction with the range datatype)

The bib-1 Completeness attribute, as well as much of the Truncation attribute, have been folded into the comparison attribute; they are replaced by anchored matching.

3.2.5 Format/Structure Attribute Type

This attribute type is used primarily to help with the interpretation of a character string operand value in cases where the comparison operator normally assumes an ASN.1 datatype; it provides guidance for the datatype conversion process. This is an enumerated or string-valued attribute. Non-repeatable.

3.2.6 Occurrence Attribute Type

Indicates the desired occurrence. For example "second subject heading".

3.2.7 Indirection Attribute Type

Indicates that the actual content of the term is not supplied, but that a pointer (e.g. url) is supplied in lieu of the actual term. This attribute will have enumerated values, e.g. URL, URN, DOI, etc. Non-repeatable.

3.2.8 Anchor Attribute Type

indicates whether a search is anchored or unanchored (see 3.2.1.2), that is, whether matching is to occur at the root of the element tree, or may begin at any node of the element tree.

3.3 Datatyping

It is recommended that term values have strong datatyping, carrying over into the definition of the comparison attributes (operators); for example, there should be separate comparison attributes for strings, numerics, etc. Groups defining specific Use attributes should consider defining ASN.1 datatypes to support their applications -- for example, personal names or dates, or geospatial information (points and polygons). There will of course be cases where the ASN.1 approach to datatyping will be too heavy-weight; in those cases the format/structure attribute type can be used in conjunction with strings to indicate that the content of a string represents some data in a specific format.

The basic datatypes defined as part of the general attribute class should include:

numerics
Integers and intUnits. These will need to be supported with the usual comparison operators equal, greater than, less than or equal, etc.
character strings
These are handled lexically. They are not assumed to contain words.
language-bearing strings
These are character strings that contain one or more words. They are treated as sets of words or phrases. This approach is felt to be better than tagging strings as "word lists", for example, distinguishing between lexical and linguistic (or at least token-based) operations should clarify queries considerably. Different comparison operators are used for language strings as opposed to IexicaI character strings.

3.3.1 Dates

There is now a Z39.50 ASN.1 Date/time definition, that should be specified when the term is a date and/or time.

3.3.2 Additional Types

Attribute set developers may define additional ASN.1 types, for example points and polygons. Personal names are an interesting "boundary" case where one might argue either for an ASN.1 based definition or a format/structure attribute indicating a normalized name according to some rules; the choice of the appropriate approach is best left to a bibliographic attribute definition working group.

3.4 Attribute Values

Although many attribute values are (and perhaps will continue to be) enumerated integers, this architecture recognizes that an attribute value may take any of the following forms:

Enumerated integer.
Integer value. For example, the value of an "occurrence" attribute may simply be the "occurrence", for example, to indicate "second subject heading" the value of the Occurrence attribute would be 2.
Character string.
A sequence of values.

4. Follow-on Actions

The ZIG should define at least two attribute sets within the new attribute set architecture (perhaps more than two; this is a packaging and granularity question). The ZIG should move away from naming conventions such as "bib-1" which imply some special legitimacy or precedence hierarchy for various attribute sets, and not use names for groups of attribute sets like "CORE". This may help avoid political debates.

One of the attribute sets (to be defined by the ZIG) within this attribute class should cover widely used basic functions, including comparison operator values, language codes, and basic expansion/interpretation values, plus query management types -- call this attribute set, for a working name, "PURPLE".

In addition, the ZIG should define a basic set of use attributes, called, for a working name, "ORANGE". in addition, a committee of bibliographic experts should be established, under auspices such as NISO, to define a new bibliographic attribute set within this general framework.

5. Other note

In version 4 the term in an operand should be replaced by a sequence of Terms. In the interim, a range of ASN.1 definition might be reserved for version 3 range comparison types, which is a pair of term values. Explicit range operators will be useful and should be added in favor of boolean combinations of operators that result in range definitions. Also in version 4, attributes on operators should be allowed.

Library of Congress
Comments February 18, 1998