The National Library of Australia (NLA) began their web archiving project, PANDORA, in 1996, and the current team consists of four members of staff. The NLA’s web archiving programme is selective, contrasting with approaches in the Scandinavian countries in particular where the aim has been to harvest the entirety of the country’s web domain. The decision to make selective harvests only was resource driven, since there was no extra funding available, although the Library are now doing periodic .au domain harvests in conjunction with the Internet Archive. .au domain harvests have been commissioned since 2005, and the 4th harvest is due this year. It will run over 4 weeks initially, and capture an anticipated 1 billion files, comprising around 40 TB of data. The Internet Archive manage the harvest and carry out a full text index; the results will be shopped to NLA and maintained on NLA servers, although copies will also be available via the Wayback Machine (without the full text indexing).
The terminology ‘PANDORA Archive’ is acknowledged to cause some confusion, particularly within the Australian government, and the Library acknowledge that they are not in fact carrying out an archive role in the traditional sense. Rather, PANDORA is a web collection, a snapshot of a point in time, a representation of what the NLA feels is important in the Australian web domain. PANDORA doesn’t meet recordkeeping needs for recording business transactions; the websites are harvested purely for their content and there is some leeway in the accuracy of dates of collection – for example, a site will be timestamped when it is harvested, but the NLA then perform further quality controls on the harvested site which may take up to a week to complete.
That said, the websites are harvested with the intention that the NLA will attempt to keep them in perpetuity, and permission is sought from website publishers for collection, preservation and public access, with this in mind. Legal deposit legislation does not (yet) cover electronic information in Australia, and considerable effort is therefore required in obtaining the relevant permissions from the website publisher (but not from every contributor). Access restrictions are applied in certain circumstances – for instance, where there is a commercial interest involved. Access can be restricted in several different ways – (a) for a set period of time following archiving, (b) for specific dates, (c) by use of authenticated logins, (d) access restricted to one PC in the NLA’s reading room. One of the problems NLA identify with their current harvesting software is that the restriction mechanism is not sufficiently finely tuned to file level – currently access restrictions can only be specified on a whole website.
The selection guidelines used are under review at the moment. The current priorities include major events (for example, coverage of Australian elections) but can basically cover any original, high quality content not available in print. The websites harvested range from academic e-journals to blogs. PANDORA is just moving into Web 2.0 harvesting, although they have already captured many blogs, some MySpace pages and some online video.
A PANDORA ‘title’ might be anything from a single PDF document to a whole or part website. A particular website might also be harvested at scheduled intervals, how long between captures depending on how regularly the site is updated, whether content is periodically removed as well as new content added, and the general stability of the organisation publishing the website. The harvest interval is re-assessed at each harvest. Currently the most frequent periods to harvest are between 6 months and 1 year. Organisationally, it is more efficient to carry out captures less frequently.
The PANDORA archive currently holds around 2TB of data, consisting of around 20,000 titles and 40,000 harvested instances.
Of most interest vis-a-vis local archive services in the UK, PANDORA has nine partners in State Libraries and other cultural organisations, who can define what they require to be collected via a web browser interface to PANDORA’s in-house harvesting tool, PANDAS. Librarians in partner institutions can also log in to fix minor problems with harvests or log more significant issues for the team at NLA to resolve. Most of the actual capture work, however, is carried out by the team at NLA.
Whilst the PANDORA team has a library background, it is noted that a certain level of technical skills are required. That said, other than the quality control work carried out on each harvested title, little post-processing is currently carried out specifically to promote the longevity of the stored files. 3 copies are created – a preservation master (the original files as harvested), a display master (which includes any quality control changes), and a metadata master. A display copy is then generated from the display master.