AustLII’s Point-in-Time legislation
system: A generic PiT system for presenting
legislation[*]
Armin Wittfoth, Philip Chung, Graham Greenleaf and Andrew Mowbray
(AustLII)
Background paper presented at the launch of AustLII’s PiT
legislation system, Sydney, 7 April 2005. For more detailed information, see the
Bibliography, examples and other information at the Point-in-time Legislation Project
page.
The point of Point-in-Time legislation
When people access legislation, what they most commonly want to find is the
current legislation on a topic. In other words, they want legislation that is as
up-to-date as possible, or (to put it another way) with all amendments included
up to the current date. The consolidated legislation produced by
Australia’s offices of Parliamentary Counsel generally sets a high
standard in the provision of such current legislation, and third party
publishers of legislation such as AustLII endeavour to republish this
consolidated legislation as quickly as possible after receiving it from the
originating source[1].
Point-in-Time legislation systems address a different problem, namely that
lawyers, judges and anyone else considering the legal implications of past
events need to know what the legislation stated at some point in the past when a
transaction occurred, or events occurred which have led to a dispute and perhaps
to litigation. They do not necessarily need to know what current legislation
states. Counsel addressing submissions to Courts, and Judges delivering
judgments are only the most obvious parties who must be as certain as they can
that they are dealing with legislation as it was at a particular point in time
in the past. Academic legal researchers who are exploring past legal
developments, or the history of a branch of the law, also need to be able to
conveniently reconstruct the history of legislation.
The conventional presentation of consolidated legislation makes the
reconstruction of an Act at some particular point in the past very difficult.
Reprints of Acts only occurred at periodic intervals, not upon every amendment,
so unless the date sought happened to coincide exactly with one of those
reprints, no consolidated version would be available. The user must then consult
the Table of Amendments (or similar) in whatever reprint is available, and
attempt to reconstruct the Act as at the correct date by going to each
subsequent amending Act and ‘removing’ amendments made since the
point-in-time of interest. It is difficult to do, often done badly, and
sometimes not done when it should be.
A user perspective on AustLII’s
Point-in-Time
Before discussing the challenges AustLII had to overcome in developing its
Point-in-Time system, we will outline the outcome of the research as an AustLII
user experiences it. From the user perspective, AustLII’s Point-in-Time
(PiT) system works as set out below. Users should remember that this system
must still be regarded as experimental, and that the 'static' or 'current'
database remains our production
system.[2] Where sufficiently
important, it should be cross-checked with other sources. Nonetheless, the
system provides an easy means to begin investigating a legal problem with a time
dimension.
When a user is browsing a section in one of AustLII’s consolidated
legislation databases for which Point-in-Time versions are available, an
additional [History] ‘button’
appears at the top of the screen. Selecting the
[History] ‘button’ takes the user
to the most recent version of the same section in the PiT system.
That section in the PiT system has three additional features:
The additional
[Versions] button (at the top of the screen)
takes the user to the ‘Available versions’ page (see below).
Additional
[Compare Earlier] and
[Compare Later] buttons (which appear when an
earlier or later version exists) take the user to a direct side-by-side
comparison of the text of the current version and the immediate prior version of
the section.
A 'Select Date' box in which the
user can enter any date, select ‘View’, and the section will be
displayed as it existed as at that date.
Whenever a user selects a particular version of a section to view, all
subsequent displays of the Act will be as at the active date of that
section, until a differently-dated version of a section is selected. For
example, if the above version of s47 for 7 July 2003 was selected and the user
then displayed the Table of Provisions for the Act, the Table displayed would be
as it existed at that date. Similarly, the [Next] and [Previous] buttons would
display those sections as they existed at 7 July 2003.
The ‘Available versions’ page (shown at right) displays each
date for which a different version of the section is available. The user can go
to each version of the section in turn, to see how the section has changed over
time. Alternatively, the user can select any two versions by ticking the boxes
next to them, click on ‘Compare’, and the two selected versions are
then displayed on one screen, side-by-side, as in the example shown on the
previous page.
Wherever two versions of a section are compared (whether via the
[Compare Earlier], [Compare Later] or
‘Compare’
buttons), the user is taken to a screen in which the texts of the two
selected versions of the section are displayed side-by-side. Where a later
version has text which is not found in the earlier version, this
‘additional’ text is displayed in blue. Where an earlier version has
text which is not found in the later version, this ‘deleted’ text is
displayed in red. The contrast between the blue and red text makes it visually
easy for users to see the differences between the two sections (often difficult
with long sections or small changes), and also provides an intuitive way for
users to understand how a section has been amended.
While the 'Available versions' page provides a convenient way to view the
history of any single provision, when the user wishes to examine various
provisions of an instrument, or the instrument as a whole at a particular date,
the Select Date Box should be the preferred option. This is because the Select
Date Box has the effect of setting the active date to that particular
date, rather than to the amendment date of the previously selected version.
Thus whichever versions were active at that particular date will automatically
be displayed without the need explicitly to select an appropriate
version.
The challenges of PiT for AustLII
Point-in-Time legislation is now a desirable part of any
‘state-of-the-art’ presentation of legislation. Developing a
satisfactory point-in-time presentation of legislation is a challenging task in
itself. However, it presents a particular challenge for a legal publisher like
AustLII that publishes legislation from multiple jurisdictions but wishes to
present the legislation in a uniform format, made searchable together, and
linked to other documents.
Free public availability implies a low cost of production, which in turn
demands highly automated processing of incoming data streams. The number of
jurisdictions covered by AustLII, and the variety of legislative instruments
from them, means that a variety of data streams, differing in format and in
quality, must be brought together, converted and presented to users in a more or
less uniform format. One of the advantages that comprehensiveness gives is the
reflexivity—between case and statute law; between different statutes; as
well as between different jurisdictions—which is a feature of
AustLII’s presentation of online legal information. Such reflexivity must
however be accomplished with minimal (human) editorial intervention. This must
be achieved without AustLII being able to exert any direct influence over the
way in which legislation is created or the streams of data that it has to work
with. In these matters it is a third party recipient like other
publishers.
In summary, any PiT legislative publishing system developed by AustLII must
be inexpensive and therefore lack significant editorial input, be capable of
being applied to any reasonable quality incoming stream of legislative data from
Parliamentary Counsels’ Offices in different jurisdictions, produce
uniform output in a format familiar to AustLII users, and produce output capable
of being linked by hypertext and searches to other sources of legal
information.
A further challenge encountered once the element of time is introduced was
that the traditional handles AustLII employs with regard to legislation
(primarily instrument title and section number) no longer remain fixed.
Fortunately, in regard to instrument titles, there exists in Australian
jurisdictions a reliable instrument handle in the form of the traditional year,
number citation (eg. NSW Act 40 of 1900). Ironically what some may have
regarded as a rather antiquated way of citing legislation reveals its utility in
the digital age.
Other publishers’ approaches
The approach taken by AustLII can be contrasted on the one hand with the
‘editorial’ approach, employed by commercial information providers,
and on the other with the integrated document production systems potentially
available to the various Offices of Parliamentary Counsel (PCOs).
Clearly some degree of automation will be employed in commercial
publishing. However, commercial publishers commonly add value to legal materials
by use of considerable editorial input (for instance writing headnotes for case
law), which is labour intensive and expensive.
With regard to the PCOs, control over standards and over document
production at all stages, from initial drafting to ultimate publication, clearly
afford advantages, and consequently ways of dealing with data streams, not
immediately available either to commercial or free access publishers. Some of
this advantage can be passed on to downstream publishers if PCOs embrace and
make their documents available in some rich, structured (and preferably
non-proprietary) standard such as SGML or XML.
In adopting their Legislative Document and Database System (LDDS), the NSW
PCO did that. The LDDS involved moving from an electronic repository relying
primarily on WordPerfect, to an integrated SGML based drafting and publishing
system. AustLII has worked closely with the NSW PCO to explore enhancements to
AustLII’s legislation based on the PCO’s new system including the
provision of NSW legislation via a ‘point-in-time’ interface. The
approach taken in relation to NSW legislation has then been applied successfully
to a substantial sub-set of Queensland and sample of South Australian
legislation.
NSW PCO has implemented its own point-in-time legislation system, by making
available as many versions of whole Acts as is necessary to accommodate changes
to any section of an Act (ie at any point-in-time). Two other Australian States
have implemented full point-in-time systems for their legislation, the
pioneering EnAct system in Tasmania[3]
and the Western Australian State Law Publisher’s
system[4]. All other Australian
jurisdictions except South Australia include some historical versions of Acts,
but these are only reprints as at selected dates, not full point-in-time systems
applicable as at any date. Overseas, full point-in-time systems have been
implemented by Parliamentary Counsel in systems including Hong Kong’s BLIS
system [5] and in
Singapore[6]. Australian commercial
publishers have as yet released few point-in time products. The TimeBase MALT
system has been applied to a limited number of Commonwealth Acts in the areas of
corporations and tax law[7] .
LexisNexis’ LawNow system includes historical versions of some Queensland
and Victorian legislation and one NSW Act. All systems are limited as yet in
their historical depth, usually only going back less than ten years, often only
five.
The legislation model for AustLII’s
PiT
The basic approach adopted by AustLII’s Legislation
Model[8] is the development of a
unitary representation of any legislative instrument over time. Once such an
historical representation has been arrived at, any particular application (such
as a web based point-in-time browsing system) should be relatively
straight-forward to develop. This can be contrasted with the type of approach
which leaves the data in the form of discrete static versions and concentrates
on developing a method of processing leading from the data to the application.
Shifting most of the processing to building a representation of the data,
rather than on handling that data, is aimed at making the system easily
adaptable to other applications, thus moving closer to the ultimate aim of
integrating case law and legislation in an historically accurate fashion.
Indeed it should be noted that beside the PiT application, AustLII’s
current production database for NSW is generated via this Legislation Model.
Moreover, the approach lends itself to multi-jurisdictional application: the
processing which handles that single representation becomes the same across a
variety of sources of input data. As described
elsewhere[9] we decided that, on
balance, a ‘structural’ approach to representing legislation would
facilitate both discovery and representation of historical changes.
The structure adopted has not needed to be overly
sophisticated[10], and in fact it
mirrors the traditional divisions of legislation across AustLII's target
jurisdictions. That is to say legislation is conceived of as an Instrument,
containing (perhaps) various Parts, in turn containing various Divisions, in
turn (invariably) containing sections or clauses. The representation of these
‘containers’ as objects constitutes AustLII’s Legislation
Model. These objects are expressed either as XML
elements,[11] or instances of
classes implemented in the Python language. XML is used to write out the
various structures to files. In turn the XML files are parsed into the Python
object model for actual
processing.[12]
Another design decision was to take the historical analysis down to the
provision level, and no further (even though sub-clausal objects are generated).
While changes could be tracked at a deeper (sub-clausal) level (especially with
high quality data such as that supplied by the NSW PCO), this decision was taken
because AustLII has always served legislation at a provision level (and would do
so in all conceivable historical applications). Viewed in the context of the
variable quality of data streams, the fact that this is how data is already
served, it should be possible to identify the components of an instrument from
any jurisdiction at least down to the provision level. This deliberate limit
might not be an appropriate design decision for some other organisation, (such
as a PCO for instance), not faced with the problem of
multi-jurisdictionality.
Approach to Point-in-Time Navigation
The system at the time of launch is intended to demonstrate and test the
core aspects of an historical multi-jurisdictional database, as well as
providing (at least in the case of NSW), useful access to our historical
legislation database. Although the depth of data for NSW (comprising nearly the
entire set of Acts dating back to July 2002) is far superior to that for the
other two jurisdictions currently included (Queensland and South
Australia),[13]the inclusion of
those jurisdictions demonstrates the multi-jurisdictional potential of the
approach taken: a generic approach to PiT.
Multi-jurisdictionality, however, comes with a caveat. Even though the
instruments are all being expressed in archive format, the quality and accuracy
of any particular database must still depend on the quality of input data. For
example (and importantly) NSW alone among the test jurisdictions supplies both a
publication and an amendment
date.[14] This means that while the
system can show changes to NSW legislation on the day the amendments were made,
with Queensland, for instance, the active date indicated states the current
published Act at that date. These are not strictly the same and users must be
made aware of the subtle differences in meaning (and accuracy) between
jurisdictions. Nonetheless the system provides a serviceable historical
interface to legislation, even if complete accuracy ultimately requires the
exercise of some human skill.
At this stage the system does not deal with commencement information, even
though this information, at least for NSW, is being incorporated into the
archive files. Commencement information is another factor where data is not
easily available for every jurisdiction. Nor does the system currently deal
with renumbered provisions or divisional elements.
What is being made available at this stage are 3 basic kinds of legislative
changes at 3 levels within legislation. These changes are textual amendments,
repeals, and additions, which can all occur at the instrument, divisional (ie.
Parts and Divisions) or provisional level. Textual amendments occurring within
the body of a provision, or to the title of a provision or divisional element
are fairly familiar. What is less well known is that an instrument itself might
change several times over the life of a piece of legislation. This creates a
problem for systems (such as AustLII's) which depend on legislative title to
identify an instrument, once the historical element is introduced. The repeal
and addition of new instruments, divisions and provisions is straightfoward. It
is interesting to note that the situation in which a provision or division is
repealed and a later distinct addition is given the same number, (a situation
which can fool extant point-in-time systems) is dealt with by the structural
approach adopted by AustLII’s PiT. In other words two separate (though
equally numbered) fragments are ultimately produced in such a situation. This
eliminates what could be a source of great confusion in studying the legislative
history of a particular provision.
The interface adopted resembles, insofar as possible, AustLII's traditional
legislation interface. An active date at which the Act is to be viewed can be
set anywhere (ie at the Table of Provisions (TOP) or within any provision) and
will affect the browsing from that point on. That is, any provision selected
from the TOP will then be viewed at the active date. Returning to the TOP from
a provision (assuming the date has been set at the provision level) will now
show the table at the new active date. Similarly selecting the Next and
Previous 'buttons' will show the next and previous provisions at the active
date. Moreover these buttons will skip over any repealed (or find any added)
provisions at that point in time (eg. selecting Next from section 2 of an act,
might at various dates find section 3 or section 2A).
Future Development
AustLII’s PiT system will be tested extensively on NSW legislation.
The Queensland and South Australian PiT databases will expand as data becomes
available in appropriate formats. We intend to next apply the PiT approach to
Commonwealth legislation, and then progressively to other Australian
jurisdictions as appropriate data, and personnel resources at AustLII make
possible.
Other Legal Information Institutes may also be interested in adopting
AustLII’s approach. For example, the Hong Kong Legal Information Institute
(HKLII) has a rich source of historical data available from the HK PCO, and
already makes available versions of sections.
The ultimate aim of PiT from AustLII’s perspective is to extend the
historical treatment to the interaction between case law and legislation, and in
particular to create hypertext links from cases to the correct historical
version of a section or Act. However, the determination of the relevant date(s)
at which an event or various events in a case took place is far from trivial, it
is difficult enough for a competent (human) lawyer to divine the relevant dates
at which to consider legislation. It will require both co-operation of the
courts and some fairly powerful heuristics to accomplish. In the short term a
‘no-later-than-this-date’ approach could be adopted, that is to say,
the law considered would have to be at least as old as the (easily
ascertainable) date of the decision. Although this might, more often that not,
find the actual law being considered, expert analysis would still be required to
find the proper date with certainty. This poses a problem for a provider such
as AustLII, which services an audience not exclusively composed of legal
experts.
More prosaically, refining some technical issues, such as intelligently
dealing with renumberings, or, where this information is available, with
commencement information, as well as the incorporation of new jurisdictions into
the system, form the immediate focus of future development for AustLII’s
PiT system.
Bibliography of Project publications
Armin Wittfoth (2004) ‘The Haymarket Legislation Model and a
Practical Point-in-Time Navigation System’, paper distributed to Annual
Meeting of Legal Information Institutes, Paris, November 2004
Armin Wittfoth, Philip Chung, Andrew Mowbray, and Graham Greenleaf (2003)
‘Can One Size Fit All?: - AustLII’s Point-in-time Legislation
Project’, 5th Law via Internet Conference, Sydney,
2001
Abstract: With the view to implementing a point-in-time legislation
navigation system, AustLII has developed a uniform schema for representing
Australian legislation. This schema has been applied to NSW and Queensland
legislation data. In this paper some of the issues encountered in implementing
such a system are discussed. Two different approaches to legislation markup,
presentational and structural, will be discussed. The importance of structural
representation to automated processing will be considered.
Armin Wittfoth, Philip Chung, Andrew Mowbray, and Graham
Greenleaf (2001) ‘Towards a Uniform Representation of Multi-jurisdictional
Legislation-like Instruments’, 3rd Law via Internet
Conference, Sydney, 2001
Abstract: As part of its involvement in a project with joint industry
partners to automate improvements to authority and interactivity to legislative
instruments, AustLII is investigating the provision of point-in-time access to
Australian legislation. The recent availability of sample data from the NSW
Legislative Drafting and Database System (LDDS) Project has prompted research
into how the legislation of various jurisdictions might be represented in a
uniform schema, thus providing a single target to the various processing tools
needed to realise point-in-time legislation. This paper discusses some of the
issues which have so far been raised.
[*] ACKNOWLEDGEMENT: This
research was supported by an Australian Research Council (ARC) Linkage Grant
entitled ‘Automating improvements to the authority and interactivity of
legislation-like instruments, to benefit government and business
performance.’
[2]For
which see out standard disclaimer at:
<http://www.austlii.edu.au/austlii/disclaimers.html>
[3] Arnold-Moore T and Clemes J,
‘Connected to the Law: Tasmanian Legislation using EnAct’, JILT 2000
<http://elj.warwick.ac.uk/jilt/00-1/arnold.html>
(as at 20/11/2001) (also published in Proceedings of Law via the Internet
1999 Conference, AustLII, UTS/UNSW, July 1999
[5] see
<http://www.justice.gov.hk/> (as at
21/11/2001). More information about BLIS can be found in Roger A, ‘Keeping
statutory law current, historical, free and useful - the Hong Kong
experience’.
[8]
Elsewhere referred to as the Haymarket Legislation Model
(HLM). [9]Wittfoth
A, Chung P, Mowbray A, Greenleaf, G, 'Towards a uniform representation of
muli-jurisdictional legislation-like instruments' Proceedings, Law via the
Internet Conference 2001, 28-30 November 2001, AustLII, University of
Technology
Sydney. [10]Implemented
as a set of XML DTDs and a near equivalent object model in Python. [11]Or
to be accurate as XML elements and schemas, for each DTD, such as update.dtd,
there is an equivalent Update class in the Python object
model. [12]For
a more complete description of the Legislation Model refer to Wittfoth
(2004) [13]South
Australia is presently converting to an XML format conforming to (but differing
subtly in the usage of) the NSW Exchange DTD, while Queensland has in 2005
converted their database to Adobe Framemaker. No conversion utility to the
update DTD has been written for the other formats
employed. [14]That
is in an easily accessible form. A survey of the Acts' notes and possibly
external amending legislation should reveal this information to a skilled human
reader.
