Dave Tarrant, Connecting preservation planning and Plato with digital repository interfaces, iPres 2010, Vienna, Tuesday, September 21, 2010 (Day 2) Session 5a Preservation Planning and Evaluation (10:30-11:50)

See also the EPrints record for this presentation including the full paper on which this presentation is based and the slides.

“impressive piece of work – excellent talk too”

timgollins #Ipres2010 #5a – @davetaz about to open the session with “Connecting preservation planning and Plato with digital repository interfaces”

mariekeguy #ipres2010 #5a @davetaz (David Tarrant) talking about Connecting preservation planning and Plato with digital repository interfaces

cardcc @timgollins well you won’t have time for tweeting if @davetaz sticks with his usual style! #Ipres2010 #5a

euanc #ipres2010: Session 5a- Preservation planning and evaluation starts. no tweets from @davetaz for the next while

timgollins #Ipres2010 #5a – @davetaz – Biggest Challenge – using and controlling the plethora of tools 2 Retweets

mariekeguy #ipres2010 #5a Tarrant: Lots of repository tools out there but repository managers unsure how to use then, need workflow

timgollins #Ipres2010 #5a – @davetaz Bit preservation is the start, then Identification, Characterisation, …. Point is useability 2 Retweets

timgollins #Ipres2010 #5a – @davetaz – gives tweet quote but I cat type fast enough !!

euanc #ipres2010 #5a @davetaz is telling us to tweet things 140 characters long and then not giving time to type them out, typical!

mariekeguy #ipres2010 #5a Tarrant: Need to lower the barriers to use of repository tools – make it fit 140 characters

timgollins #Ipres2010 #5a – @davetaz Talking about e-prints – already at Torrent speed and speeding up

timgollins #Ipres2010 #5a – @davetaz Droid profile of a publication repository – thanks for the Plug Dave

mariekeguy Talks faster than me! RT @timgollins: #Ipres2010 #5a – @davetaz Talking about e-prints – already at Torrent speed and speeding up  1 Retweet

euanc #ipres2010 #5a @davetaz speaking too fast for my slow brain this morning. impressed by @mariekeguy & @timgollins tweet speet

timgollins #Ipres2010 #5a – @davetaz E-prints – already missed the point about Characterization using JHOVE through Plato

“Plato action plan XML read by Eprints for ‘action’. V.cool.”

euanc #ipres2010 @davetaz : only get factual information in registries, not institutionally subjective information

timgollins #Ipres2010 #5a – @davetaz Now on Risk assessment – risk is local – do the risk analysis – central registries nly have global risk factors

euanc #ipres2010 we need a tool that identifies intellectual entities that are made up of multiple computer files via indicators to be defined

timgollins #Ipres2010 #5a – @davetaz e-prints talking about sampling for characterization and then using Plato for Risk assessment 2 Retweets

euanc how do you test results of bulk migration? how do you confirm you have preserved whatever you were trying to preserve? #ipres2010

timgollins #Ipres2010 #5a – @davetaz – talking about executable preservation action plans that use linked data registries to identify tools 1 Retweet

timgollins #Ipres2010 #5a – @davetaz Plato – executable action plan contains decisions taken in creating the plan – preserve the plan in the repository 2 Retweets

timgollins #Ipres2010 #5a – @davetaz Trust, authenticity – declare the plan used to migrate the objects – transparency/provenance delivers trust ! 2 Retweets

“Good to see KeepIt project enabling Repo managers to really use tools”

euanc #ipres2010 #5a @davetaz – quotes from KeepIT exemplars “Much more time-consuming and complicated” “less confident but now knw what it means” 1 Retweet

mariekeguy #ipres2010 #5a Tarrant: KeepIt examples – repository managers realise dig pres not just for the techies – also part of their role 1 Retweet

neilgrindley @davetaz talking about integrating preservation tools. Good to see KeepIt project enabling Repo managers to really use tools.#ipres2010 1 Retweet

mariekeguy #ipres2010 #5a KeepIt project http://bit.ly/92Qhf1 1 Retweet

timgollins #Ipres2010 #5a – @davetaz Presenting the e-prints users response – impressive piece of work – excellent talk too  2 Retweets

pjvangarderen #ipres2010 Tarrant: Plato action plan XML read by Eprints for ‘action’. V.cool. Need to see if this will work for Archivematica. 2 Retweets

jisckeepit Thanks @timgollins @euanc @mariekeguy for excellent commentary and summary on @davetaz Plato and repositories #ipres2010 #5a

Questions?

The short question and answer session that followed the presentation is summarised by the presenter, Dave Tarrant:

Q1: How did you organise the training courses in KeepIt and what materials did you use.
A1: We utilised connections with projects to invite key people to give presentations which were focused on repository managers, we collected the materials and these are available online (url tweeted http://www.ecs.soton.ac.uk/research/projects/640, source materials). Ed. Or try course blogs, course presentations (Slideshare)

Q2: Is there any format you can’t handle?
A2: In EPrints, no, in the scope of the tools, yes, as the tools do not cover all formats. However in the case of eCrystals they have started this process by providing the identification of their files to PRONOM-DROID. You can then follow the workflow presented to look at extracting characteristics and feeding these to a tool or developing a new characterisation tool.

Q3: I notice in the paper you give the plan an ID (ID/41). Is this a persistent ID?
A3: Yes! An eprints persistent ID is assigned to almost everything in the system. I simply removed the first part of the URI to fit it in the 2-column format. EPrints provenance (which uses the Open Provenance Model, OPM) wouldn’t work without it.

Summary

So what can we divine from this? The presentation was fast and furious, in true Tarrant style, but thanks to the tweeters we can see the main points were mostly effectively conveyed and the story well received and complete. An amazing response.

lescarr Following #ipres2010 where @davetaz presents the EPrints/DROID/Plato integration so quickly that tweets are too long & cumbersome to keep up

So far in the practical work in KeepIt course 4 we have managed some storage services from our EPrints repository, then deposited some image files in the repository and performed preliminary format identification and risk analysis. We exported those files and used Plato to produce a preservation plan for that format.

These tools were only trivially connected by the export and use of common files. In other words, you could have done these courses on EPrints and Plato independently. Further, since the stated aim of this course module was to put preservation in the repository interface, you might ask how this is achieved when all we have done so far is copy content from the repository and instead used Plato.

Here is the clever part …

and it’s simple compared with the previous session. Recall the repository interface when we exported our GIF files. There are no presentation slides for this final session in course 4, just instruction sheets for our remaining practical work, so an illustration of this interface is reproduced here.

Under the preservation actions button we used to download and export our GIF files to Plato is another button, to upload a preservation plan. This is the plan we have just produced and saved from Plato, as an XML file. So select that file and with one click we are back in the repository, with a preservation plan. In this case our simple plan chooses to migrate GIF images, considered high risk (hypothetically) to PNGs, our choice of low-risk (hypothetically) image formats.

We can now follow the instruction sheets for this session to complete our work for this module.

• Actions: find the original GIFs, upload the plan, view the result.

Each image record now contains a migrated version, the PNG, as well as the original GIF.

We can now update our earlier format risk screen, which shows the new PNGs as low risk objects and includes our GIFs in the same category, because they have migrated versions, but with a red bar to indicate they were originally considered to be high risk. Below this screen is shown the list of preservation plans relating to these objects.

A further short exercise demonstrates that when we deposit new objects in this format, we don’t have to reload the preservation plan. The appropriate plan is recognised and an Enact Plan button appears for our high-risk objects.

The exercise completes with a short demonstration of provenance. As we learned in KeepIt course 3, the provenance of an object is a verified record of its past history. Since our image files have been converted we have a provenance record for them. In EPrints this relational information is stored as Linked Data, and in this exercise we learn how to view these relations.

EPrints-Plato KeepIt course 4 summary

We said at the outset of KeepIt course 4 that we would put preservation in the repository interface, and we have done this for storage, risk analysis and preservation planning. We believe this is the first time this range of preservation support has been available in a major repository platform.

We also said that while you could do courses on EPrints and Plato separately, this is the only course to combine the two. We have shown here how the value multiplies by combining the different tools.

Versions of this EPrints-Plato course have been given in Corfu (ECDL, Sept. 2009) and Madrid (OR, July 2010), as well as this one in Southampton as part of the KeepIt course, and was just presented for the final time in Vienna (iPres, Sept. 2010).

It remains to me to thank our outstanding presenters for these courses: Andreas Rauber and Hannes Kulovits from PLANETS and Vienna University of Technology, and Dave Tarrant and Adam Field from the EPrints team at the University of Southampton.

Preservation planning provides a workflow leading to a preservation plan. One of the problems with some approaches to digital preservation is they are too proactive and reactive where file formats are concerned. For example, take a file format, decide the consensus is against it (such as Microsoft Office formats), and migrate. There is very little formally to justify this process, just that it is possible and not very difficult. The longer-term consequences of this action are unknown. What was done with good intent may turn out to be detrimental.

Preservation planning seeks to give a more formal basis to such decisions, and in the process will help to automate and record the consequent actions.

For this session in KeepIt course 4 we welcome back Andreas Rauber and Hannes Kulovits from the Vienna University of Technology to provide an extensive presentation on preservation planning using a tool, Plato, which they developed as part of the PLANETS project. This will be followed by further practical work.

As a formal process, preservation planning requires, unsurprisingly but perhaps disconcertingly, a lot of preparation as it takes account of preservation policies, legal obligations, organisational and technical constraints, and user requirements (slide 8). Fortunately, participants in this course are well positioned to do this, since we have covered many of these issues already in KeepIt courses 1-3.

The preservation planning approach supported by Plato can be overlaid on the OAIS reference model (slide 10), and is shown in more detail in slide 12. Preservation planning with Plato involves four stages:

• Define requirements (slides 12-33)
• Evaluate alternatives (slides 34-45)
• Consider results (slides 46-58)
• Build preservation plan (an exercise)

The reader can explore the Plato workflow using the slides. Here we will highlight some of the critical stages.

In the KeepIt course, and indeed throughout the KeepIt project, we have begun our preservation approach with format identification, but now we go further. We have to relate our identification and other information about our digital objects to our requirements for those objects. This is where our understanding of the significant characteristics of digital objects from KeepIt course 3 becomes useful. Slide 25 shows a mindmap of the sort familiar, again, from course 3, and in slide 26 we encounter the Plato interface, the tree editor, for the first time. Flipping between mindmap and Plato, the following slides show how the requirements are elaborated and values added, illustrating how this information might be mapped to the Plato editor.

Once we have described our objects we want to know what we might do to preserve them. There is typically more than one choice, not just in terms of a preservation action (e.g. format migration) but also in how that action might be performed, and these alternatives need to be evaluated to serve our requirements. If you recall, at the end of the previous session in this course module, we deposited some GIF files in a test repository and then downloaded those files in readiness for this session. In slide 37 we see these files appearing in Plato for the first time. Plato now shows us what alternative actions are available for these files.

Now there is a decision to be made: go/no-go/deferred-go. To make an informed decision we need to run some experiments, that is, to run the alternatives and compare the results before we commit to any plan. Plato helps us to run and evaluate the experiments, in this case on our image files.

Having begun to get some results we could perhaps begin to think we have done the hard work, but there is still a tricky stage to negotiate. Before we can analyse the results we have to transform and weight the measured values from the experiments, that is, to normalise the values so different experiments are measuring the same thing (slide 48-49), and to set the level of importance for each of the factors in our requirements tree (slides 51-52).

Finally, Plato presents our results (slides 55, 57) and we can see the benefit of using this tool.

Summary and exercise

Before we start the exercise, slides 60-64 summarise the presentation so far. If there is one conclusion that I would highlight above the others, it is that

• preservation planning is a basis for well-informed, accountable decisions (slide 64)

It is no longer necessary or acceptable to make ad hoc preservation action decisions. This has been a detailed and involved process, but the benefit for a large repository is that the resulting plans can be used across all content in the analysed formats, now and in the future.

Two exercises are set out in slides 66-70. Again, we use our imported GIF files. My impression from observing participants on the course was that this may have been the hardest exercise in the whole KeepIt course, especially exercise 1 which confronts the requirements, and the first encounter with the tools, including the Freemind mindmap tool. There is nothing like a steep learning curve to get the best out of people, and by the end of this session you could hear the sound of pennies dropping.

We now have a preservation plan, and in the next session we will put that plan to work in a repository.

DIGITAL PRESERVATION is

NOT so DIFFICULT

if you WANT to DO IT

You will want to do digital preservation if …

View the conditionals now, see slide 2 in the slideshow below, or read on.

Thus began the introduction to my brief interlude linking two practical sessions in Dave Tarrant’s 90-minute EPrints Preservation workshop at the EPrints User Group meeting, at the Open Repositories 2010 Conference in Madrid. The workshop aimed to connect preservation planning with tools provided for use with EPrints repository software. My role was to say something interesting, preferably on the theme of digital repository preservation and, if we were lucky, to link seamlessly back to the second part of the workshop, to the extent that participants would be refreshed, ready for the next challenges, and will know and understand a little bit more about what is to come and be better prepared for it. In this talk we also considered the role of file formats and the essentials of preservation workflow and preservation planning.

The primary resources for this workshop include, specifically, the File Formats exercise (short version) preceding this presentation, the Action and Provenance exercise following it, and the main Presentation. The workshop was scheduled to last 90 mins, so independent users can expect to gain something from these materials in a similar or less time.

Notes are provided with these presentation slides and can be found by using the View on Slideshare button (bottom right-hand corner on the slide viewer above, or try here). Slideshare seems not to have reproduced the table in slide 12; the original can be found as Table 2 in this earlier blog entry.

@jisckeepit For the few at #or10 not at the EPrints preservation workshop (cough), a quick summary follows in three tweets
10:57 AM Jul 9th

Preservation institutions have identified a workflow for managing file formats and built tools, but no joinup and common interface
10:58 AM Jul 9th

So build preservation tools such as DROID, Plato, etc, into EPrints and access through common repository interface
10:58 AM Jul 9th

The clever part, which our workshop participants now know, is the button for importing a preservation plan from Plato to EPrints
10:59 AM Jul 9th

More on digital preservation conditionals

Originally I had not included ‘so’ in my opening statement. Instead it read: DIGITAL PRESERVATION is NOT DIFFICULT. When I arrived at OR10 the first item I fished out of my delegate pack was a colourful cardboard flyer for DuraSpace, and on the front it clearly said: “Preserving the world’s intellectual, cultural and scientific heritage isn’t easy”. I had to admit they have a point, and I was merely sloganising as a the simplest means of reassurance. So I modified my claim, but the key part really is in the qualifiers.

We have to understand why digital content managers and repository managers are concerned about digital preservation, yet why this translates so little into practice outside specialist preservation institutions. (And to understand the turnout at events like this: @jisckeepit Maybe ‘last day-preservation’ not a perfect fit. So thanks to those who turned up. We hope you found preservation works with EPrints #or10 10:06 AM Jul 9th.) To gauge at what point digital content and repository managers might expect a natural transition from interest/concern to practice, we produced this rough rule-of-thumb metric. If one or more of these criteria apply, then the application of digital preservation is likely to become magically less onerous and more beneficial for your content. So here are the conditionals:

DIGITAL PRESERVATION is

NOT so DIFFICULT

if you WANT to DO IT

You will want to do digital preservation if you have

1. a lot of digital content
2. collected over years
3. a specified responsibility and resources for that content
4. an understanding of how that content is used now
5. how it will be needed in future,
6. how the type of content you collect may change going forward

Can you add more?

Among our KeepIt exemplar repositories, I would say at least three can apply point 1, while perhaps only one might say that point 2 applies, so far. Due to their status as ‘institutional’ repositories, all four exemplars would have understood point 3 but as a result of participating in the project are likely to understand better the connection with preservation, and we hope that all four are making progress on points 4-6 as a result of the KeepIt course. So the conditions apply broadly, to include different types of repository at different stages of development. They do not exclude any repository, and can act as a rough indicator of when preservation should become a higher priority that ought to be properly resourced.

If one or more of these six points describes your repository, you are ready to act now. There are tools available to help. Real tools for real content. Now those tools can be applied directly to your repository – for preservation planning and workflow, and for storage – all accessed and controlled through your repository interface.

Whether your content includes text documents, images, sound, vision, science data, or is used for teaching or research in science, arts or any other field, there are strategies to help you.

If that doesn’t describe your repository now, focus on how you plan to get there. Getting to digital preservation will be your success story.

For some repositories there are preservation tools built into your workflow. For others there are just tools. You can work it out.

UAL Research Online is a specialist repository of research outputs in arts, design, and media, operating on a version of EPrints that has been customised to be able to hold, manage and showcase our mainly practice-based research.  The research outputs of our university (University of the Arts London, which consists of London College of Fashion, Central Saint Martins College of Art and Design, London College of Communication, Chelsea College of Arts and Design, and Camberwell College of Art) are rarely text documents. They are exhibitions, paintings, textile designs, events, stage designs, films, costume design,  sound art, industrial designs, photography, sculpture, installations, etc. This means that our institutional repository is rather different than any other.

Our file formats include:

Images: jpeg, png, bmp, tiff, gif, pdf
Audio: avi, mp3, mpeg4, wav, ac3, flac, ogg
Video: mov, mpeg, quick time, flash, avi, theora/ogg

We are also beginning to include archived websites.

Because of this diversity, our preservation issues are a little more complicated. It will be important for us to use the EPrints extensions (developed by Dave Tarrant of the University of Southampton) that incorporate format recognition, and we will upgrade to the version of EPrints (v3.2) which these tools require, before the end of the summer of 2010.

In addition to implementing the tools developed within the KeepIt project, based on the KeepIt course modules of the various preservation tools available, I have chosen to work through the online preservation tool DRAMBORA, as it best suits the needs of UAL Research Online at this point in its evolution.

I chose this tool from among the many we discussed during the KeepIt course for the following reasons:

it is designed for repositories rather than all digital assets of an organisation;
it can be applied to very new repositories;
it is a self-assessment exercise;
it does not require advanced technical knowledge

DRAMBORA stands for “Digital Repository Audit Method Based on Risk Assessment”. It is sponsored by JISC and managed by the DCC, the Digital Curation Centre in the UK.

DRAMBORA defines digital curation as the management of risk. The repository manager establishes the objectives, activities, and assets of the repository, and then assesses the areas of risk – identifying weaknesses and strengths, and then managing the areas of risk.

Essential to DRAMBORA’s approach is the belief that “the job of digital curator is to rationalise the uncertainties and threats that inhibit efforts to maintain digital object authenticity and understandability, transforming them into manageable risks.” DPE Newsletter, Issue 2: September 2007, p.9

DRAMBORA includes the following steps:

• Defining the mandate and scope of functions of the repository
• Identifying the activities and assets of the repository
• Identifying the risks and vulnerabilities associated with the mandate, activities and assets
• Assessing and calculating the risks
• Defining risk management measures
• Reporting on the self-audit

After the DRAMBORA exercise is completed, UAL Research Online should have:

- a ‘comprehensive and documented awareness of mission, aims, objectives, activities and assets.’

- a ‘catalogue of pertinent risks, categorised according to type and relationships, which have been described in terms of ownership, probability and impact’

- ‘internal understanding of shortcomings of the organisation – so that resources can be allocated or redistributed to pressing areas’

We should also be prepared for an external audit, if needed. Compatible external audits are said to include:

- Trustworthy Repositories Audit & Certification (TRAC) – an accreditation of the US National Archives and Records Administration,

- Nestor Catalogue of Criteria for Trusted Repositories, or

- Consultative Committee for Space Data Systems (CCSDS) digital repository audit assessment criteria

One of the objectives of UAL’s participation in the KeepIt project  (as defined by my predecessor as manager of UAL Research Online) was to write a series of guides for digital preservation, meant to advise staff about it and to impress the need for it to our senior management. I hope that the DRAMBORA results will feed into this document as well.

So I explored the DRAMBORA site and signed up for the process. I have completed the first stage, in which I defined the functions and scope of the repository. Already I found much food for thought and have several questions I need to ask the senior managers about the specifics of my repository mandate – I can see that DRAMBORA will require me to think through more than just preservation risks, and will be helpful in specifically defining other aspects of our repository.

After I’d done this, we were fortunate to have a visit from Martin Donnelly of the Digital Curation Centre at the fifth module of the KeepIt course, which was held on 30 March at the University of Northampton. Martin gave us a thorough grounding in DRAMBORA and we were able to complete some practice exercises. Interestingly, at the end of the course we were polled for our reactions, and all 15 respondents indicated that DRAMBORA either could be useful, they intended to use it, or they have used it; no one was unlikely to use it.

Martin advised us that our audit scope and purpose must be decided ahead of time, and we must make it clear at which stage of repository development the audit is being performed. It’s important to realise that no repository exists in a vacuum – we are embedded in our institutional management structures and policies, as well as the limitations and possibilities of our IT support provision, the climate of research we function in, and the wider world of UK higher education and funding.  We need to be clear on the repository’s goals: What do we do/What will we do?

Another of the important preliminary steps Martin highlighted was the need to ascribe selected “functional classes” to the repository – for example, metadata management.

We had a workshop session in which we filled out a sample section of the assessment (for reference, this was Stages 4-5-6 on Form T8/T9/T10). Our group looked specifically at T10, entitled Manage Risks. The form asked us to name and describe a possible risk, and then explain its manifestation. Then we classified the nature of the risk, identified the risk owner and stakeholders, listed the risk relationships, probability, potential impact, and from these we calculated its severity. Then we devised a risk management strategy, a risk management activity, and identified the owners of these two. This was a lot of work! and much group discussion ensued. It was a bit difficult to do as a group, because we found our repositories were all quite different, even in terms of the sorts of risks we each thought we’d be likely to face. But it’s clearly a very thorough process.

After trying T10, I  was apprehensive about my ability to think up all the possible risks that the repository faces, but was glad to learn that the DRAMBORA pdf guide includes lots of examples of risks repositories may face. The DRAMBORA website claims: “DRAMBORA Interactive provides a host of real-world risk exemplars which you can use or modify for your own repository’s circumstances.” I think this is a crucial part of the process, and I’ll certainly need to refer to these examples.

A minor concern the emerged during the hands-on experience with T10 is that filling out a lot of these forms will be tedious – I envision that there will be a lot of repetition, e.g., stakeholders will be the same for many risks.  Also, although the fact that DRAMBORA is a self-assessment is one of its good points, I do wonder if I am qualified to assess all these areas.  I don’t see how to independently test my decisions, so how will I know the probability of x happening? DRAMBORA is meant to ‘provide peace of mind’ but if it is based only on my judgement, I wonder how reassuring it will be.

Martin advised that we should allot 5 full working days to the self-audit, and I am not sure where I will find this much time to devote to this, despite my best intentions and awareness of my responsibility as a KeepIt project partner and exemplar.  I will have to put it together in bits and pieces, rather than get immersed in the task for a block of time; the latter would be far preferable. It was suggested that there might be a possibility of a ‘DRAMBORA Light’, that I could put together for myself and report on, for the use of other repository managers that are as busy as I seem to be. There are lots of exciting things going on for UAL Research Online in the next months, including our EPrints software upgrade, the complete restructuring of the university’s research office, adopting the repository to be used for all research reporting functions in the university, and my involvement in three additional projects with their own sets of deadlines, meetings and papers to write. It’s easy to keep putting off getting properly stuck into DRAMBORA, and it’s not just about my own time management – I think this illustrates a common problem for digital preservation generally. We all know that we very much need to assess, manage and minimise risk, but preservation tasks tend to fall into the ‘Important’ category, not the ‘Urgent’ one. It’s easy to spend six months attending to ‘Urgent’ work, and never get to any of the ‘Important’ bits.

Over the last few months the need for a good risk management has been very dramatically demonstrated in the news – I wonder if it would help to post this photo in a prominent place in my workspace?

Deepwater Horizon oil rig fire
Photo courtesy of the U.S. Coast Guard Eighth District External Affairs http://www.flickr.com/photos/uscgd8/4542934710/
Licensed under Creative Commons Attribution/Share Alike 2.0.

Stephanie Meece
Institutional Repository Manager, UAL Research Online
University of the Arts London

So far, in the KeepIt training course we have been introduced to a series of tools that will help us, the repository managers, to prepare our repositories for the long term preservation of their content. These tools have covered aspects of organisational preparedness through strategy and policy (DAF and AIDA); issues around costing (KRDS and LIFE3); and description for preservation using significant properties, metadata and provenance (InSPECT and PREMIS).

In session 4 of the course we finally reached what I believe will be the core tools for repository managers – the Eprints preservation apps (including the storage controller) and the PLANETS tool, Plato. Although I’ll be concentrating on Plato in this post, it will really be the interaction between Eprints and Plato that I hope will allow me to preserve my repository content in a manageable and cost-effective way.

About Plato

First released in November 2007, Plato is described as a ‘preservation planning tool’ . It defines a consistent workflow which will lead to a complete preservation plan for a given set of objects:

“A preservation plan defines a series of preservation actions to be taken by a responsible institution due to an identified risk for a given set of digital objects or records (called collection).
The Preservation Plan takes into account the preservation policies, legal obligations, organisational and technical constraints, user requirements and preservation goals and describes the preservation context, the evaluated preservation strategies and the resulting decision for one strategy, including the reasoning for the decision. It also specifies a series of steps or actions (called preservation action plan) along with responsibilities and rules and conditions for execution on the collection. Provided that the actions and their deployment as well as the technical environment allow it, this action plan is an executable workflow definition.”
http://www.ifs.tuwien.ac.at/dp/plato/intro_documentation.html

Our tutors, Hannes Kulovits and Andreas Rauber of the Vienna University of Technology, took us through the various steps in the development of the preservation plan:

Starting with defining requirements, we considered the context of the preservation plan, including what triggered the preservation planning activity. Institutional constraints, legal obligations and the target community all play a part here. If the organization has a preservation mandate or mission statement then that is relevant at this stage too.

For the purposes of the plan, a selection of records need to be considered. The sample should be representative of the features and characteristics of all objects in the collection. Stratification by file type, size, content and time of creation may be appropriate. We were advised to use the DROID and JHOVE tools to identify file formats.

To identify requirements it was stressed that we would need input from a wide range of colleagues, including content producers, managers, lawyers, technical specialists and others. The purpose of this step is to define all the relevant goals and characteristics of the plan. Four groups of characteristics were suggested: object characteristics, record characteristics, process characteristics and costs. In the practical exercise for this section we used the Freemind mind-mapping tool to describe the linkages between these characteristics and then we used the in-built facility to import the requirements into the Plato tree editor.

Rather than try to describe the full set of requirements, KeepIt course members each tackled a small part of the requirements tree. Even this was sufficient to cause much discussion among our small groups – especially when it came to assigning measurable units to each ‘leaf’ of the tree. In our group we decided that a set of templates showing the ‘normal’ requirements for a range of object types would be a useful addition to the Plato tool. This would give new Plato users a benchmark against which they could consider the specific requirements of their own institution.

Moving on to the evaluation of alternatives, we thought about the suitability and possibility of different preservation strategies and tools for each object in the sample. Migration and emulation are the most obvious contenders. Examples of alternative strategies might be conversion from DOC to RTF or PDF format, or migration from one version of PDF to another. For each alternative it is necessary to define which tool to use, which functions and parameters of the tool, and what resources would be required. This leads on to the Go/No-Go decision for each alternative: whether to continue the preservation procedure or not.

If continuing, then the next tasks are to develop, run and evaluate standardized experiments on the object. The PLANETS testbed was used for this. The value of the experimental approach is that objects can be moved through a consistent set of steps, producing results that are comparable and repeatable. By conducting experiments, different tools can be evaluated and the outputs assessed according to the requirements previously defined. The most appropriate tool may then be chosen for the eventual transformation of the objects in the collection.

In our group exercise we experimented with using different tools to convert image files from .gif to other image formats. We examined criteria such as the availability and ease of use of the tool, the change in file size and image quality and the time taken to perform the transformation. In real life, each of these criteria would be weighted according to its relative importance to the organization and these weights would be allowed for in the analysis of the results.

Having followed the Plato process, it is easy to have confidence in the resulting preservation plan. The methodology is both thorough and sound, and decisions based on this will be fully accountable.

Using Plato in the repository

Like most tools designed to support digital preservation, the Plato tool was not originally intended for use in repositories. Repositories are often complex digital collections containing files in a multitude of formats, with metadata that may be inconsistent and/or incomplete. So the Plato concept of ‘collection’ is potentially very helpful. It enables the repository manager to address the preservation needs of a subset of repository content, a ‘collection’ defined by a set of common characteristics. For each collection a preservation plan can be created and then implemented as needed.

Recognising the point of need is where the new tools in Eprints come in. Eprints is now able to perform an analysis of file formats and identify those that are at risk of no longer being accessible or editable. This information can be used to trigger action according to the preservation plan created in Plato. The transformed objects can then be re-imported back into Eprints, now at a much lower risk of loss.

Of course this does not obviate the need for vigilance by the repository manager. In the discussion which followed the presentations on Eprints and Plato, Dave Tarrant reminded us to be proactive about preservation – to identify potential risks and make repository users aware of these. For some repository content there may be no migration solution (e.g. for scientific datasets); the repository manager may have to make the risk explicit (e.g by documenting it) and allow others to develop a preservation solution.

Nor do the tools provide all the answers. Plato and the new Eprints tools are both in a relatively early stage of development. As Andreas said, showing that a prototype works is quite different from widespread deployment. These solutions need to be turned into a preservation infrastructure, supported by robust digital preservation standards.

Cornell University three-legged stool model for digital asset management

KeepIt course module 1, Southampton, 19 January 2010
Tools this module: DAF, AIDA
Tags Find out more about: this module KeepIt course 1, the full KeepIt course
Presentation referred to in this blog entry The AIDA toolkit: Assessing Institutional Digital Assets (Slideshare)
Presentations and tutorial exercises course 1 (source files)

Steve Hitchcock has organised a series of modules on Digital preservation tools for repository managers, and I was invited from ULCC to present something on Assessing Institutional Digital Assets (AIDA) at the very first module, on 19th January. It was very good to meet with such a warm reception from the intelligent and lively audience, many of whom were repository managers themselves, or involved with building and developing repositories. It was clear from the start they were all very engaged with the work, and understood the issues well.

I thought it would be interesting to set them an exercise that explored two data-management activities within the AIDA toolkit, namely ‘Metadata Management’ and ‘Access and Sharing’. The AIDA self-assessment toolkit is intended simply to offer a snapshot of an Institution’s readiness to carry out management of its digital assets, assessing that capability across three ‘legs’ – Organisation, Technology and Resources – while applying the assessment at the level of the entire Institution, and of a single repository. AIDA’s proposition is simple – the assessment will almost always result in a wobbly three-legged stool, quite often showing that the technology leg is the most advanced of the three. Steve pointed out that a result like this need not surprise us, and this is especially true given the advances being made with a tool such as eprints.

Concentrating on two activities from the larger and more complex AIDA framework was instructional – for me as well, as the manager who must put work into the toolkit to improve it. Two ambiguities in the metadata strand were exposed by the keen minds of the Southampton audience; did AIDA refer to discovery metadata, technological metadata, or preservation metadata? More importantly, the written exemplars in AIDA seem to be making an assumption that automation of metadata is a commonly-desired goal, and that the results of automation are always good. But some repository owners were proud of the quality of their “hand-crafted” metadata.

However, in the closing minutes of the day, one team fed back to me on their discussions on the Access and Sharing strand. Their AIDA-based deliberations showed to them quite clearly that their test repository scored highly in the organisational leg (managed shared storage, centralised management, and agreements about cross-department sharing) and the technological leg (technological capacity and an appropriate infrastructure to support those policies, and how well the strategies are aligned), but there was an imbalance in the resources leg, and resources allocated to technological development were not quite at the correct level to match the strategies and policies. This isn’t simply a matter of lacking the money and staff (who doesn’t!?), but a simple graphic demonstration of where the stool is wobbliest, through findings which can be backed up through the provision of documented evidence, and one which might enable the repository to take steps to achieve stability on all three legs. Many heads in the room nodded instantly as they recognised themselves, and one comment was “I think we could say a lot of repositories fit that model”. A result like that is pure gold to me as the AIDA owner.

From TIFF to JPEG 2000? Preservation Planning at the Bavarian State Library Using a Collection of Digitized 16th Century Printings

Although ostensibly about two particular image file formats, the real point of interest for us is how this acts as an example of preservation planning. This is a topic that we will encounter later in our project training, to be presented by Andi, and based on the preservation planning tool Plato, also described in the paper.

Preservation planning connects the processes of managing file formats within your collection. Computer applications change, so some file formats are at risk of becoming obsolete, and when this happens the content may become inaccessible. To prevent this preemptive action might be taken, but you have to know where (for which files or contents) and when such action might be taken.

The first step in this process is to identify the formats of all content files within a collection. Then you have to know the preservation implications of each format, and decide on appropriate actions, if any. This is where it gets tricky because the the number of risk factors to take into account for each format is large – status of applications and viewers, etc. – and thus making decisions on action becomes more complex. Bear in mind as well that every preservation action incurs a cost, and costs can rise rapidly for large, diverse digital collections. So good judgement is critical. By connecting these steps, preservation planning can help produce good judgement.

The D-Lib paper provides a practical example, but note that where this paper evaluates the migration, a preservation action, between two formats for a large volume archive, the exemplar repositories in KeepIt and other institutional repositories typically switch these factors around, i.e. smaller collections but with many more formats and consequently multiple possible conversions. I’ll leave you to judge the effect on complexity caused by each factor.

Now the really sobering part.  The result of a preservation planning process, say Rauber et al., is dependent on the institutional and repository context: “the design of one plan can differ considerably from the plan of another institution, even one with a similar collection.” Considerations include “the institution’s preservation policies, legal obligations, organizational and technical constraints, requirements, and preservation goals, as well as the capabilities of the tested tools. Preservation Planning is a process that depends very much on an institution’s individual policies and requirements in its day-to-day work.” In reality, there is no magic wand.

This is where Plato comes in, and this is why we are trying to integrate it with EPrints, so that at least you might be able to invoke the planning process within a familiar repository environment. It’s also why we are not starting the training programme with preservation planning.

Finally, the conclusion of the paper highlights another factor, time dependence. The current recommendation for the image example, despite positive reports of the target format JPEG 2000 and all the accompanying analysis, is not to migrate. But this could change: “in one year we’ll look at this plan again to see if there are more tools available and whether or not the ones we considered in this year’s evaluation have been improved.” In other words, planning is an ongoing process. There is no single result, and even that can change over time, all depending on your repository, of course.