Originally posted at blogs.it.ox.ac.uk on April 6, 2012 by Martin Wynne. IT Services at the University of Oxford has decided to delete a large number of historical blogs, and this is one of a number of posts related to the Oxford Text Archive which are being re-published here, after being laboriously retrieved from the archive provided by the Wayback Machine.
The following is a partial summary of a presentation given at the Interedition Symposium in the Hague in March 2012 on the topic of Scholarly Digital Editions, Tools and Infrastructure.
People are often talking about digital silos in the context of digital resources in the humanities. The problem is that resources, although valuable in themselves, are located in different locations on the web, where they might be difficult to find, and they all have their own individual interfaces and registration procedures, and are not connected with similar or related resources. So you can’t easily search the Old English Corpus (available either for download with no software from the OTA, or online via numerous university library portals to local users). Some resources, like the ARCHER corpus, you can’t access at all unless you’re friends with someone at the University of Manchester.
This is clearly far from ideal. But what alternative, more connected, architectures are most appropriate to achieving interoperability and sustainability of the arena of digital textual scholarship? The emergence of fast and high capacity networks, a deluge of data, and web service APIs mean that it is increasingly possible to imagine and build distributed architectures for scholarly services, where data, tools, computing resources, and the outputs of annotation and analysis live in different parts of the network but can be brought together virtually in the user’s desktop environment. The current concerns about ‘digital silos’, in which the outputs of digital humanities projects are deployed online unconnected to other resources, and with limited sustainability, are directly addressed by this vision.
I want to put forward put the argument for distributed architectures, while reviewing some of the risks and problems, and survey some current moves towards such an infrastructure. And I also want to suggest another metaphor as an alternative to the ‘silo’.
An open and fully distributed architecture where the resources are located in different places can have the advantages of allowing the following services to be created:
- potentially unlimited functionality, since developers can deploy content and tools that they want to use, and which can interoperate with other data, tools and infrastructure services;
- building ad hoc collections and corpora across different repositories;
complex workflows, for example piping together web services from different locations;
- protected resources (e.g. works in copyright, sensitive data) curated in situ yet still analysed online via web applications which access the data via a secure infrastructure.
All of this can happen in a situation with a better division of labour than we typically have now: the repositories don’t have to worry about tools; tool and content developers don’t have to worry about creating the entire online environments; tool developers don’t have to worry about data management; users don’t have to install software; etc.. The emergence of an ‘ecosystem’ with numerous actors providing content, tools, computing resources, and other infrastructure services, provides a flexibility and resilience and the potential for sustainability which is not possible for a single-site or other more closed or monolithic system.
So let’s consider the unconnected, problematic online resource as a fishtank rather than a silo.
There are lots of fishtanks out there, and they can be very large, elaborate, pretty, sophisticated, long-standing and sustainable. But they’re all in different places and they are not connected with each other. If you want to see a variety of fish, you have to visit a lot of houses, try to negotiate access to their fishtanks, and make use of whatever facilities they have for viewing or otherwise analysing the fish. Some places are better than other to visit – aquariums might have very good facilities and lots of information, but you still can’t view the fish in one aquarium alongside the fish in another, and it’s hard to compare them.
And if I want to keep a fish I have to build a fishtank and maintain a fishtank, or I could find someone else’s fishtank to put it in, but then it’s difficult for me to get access and control the environment. And who’s going to carry on feeding the fish? We can probably agree that it’s better if we don’t all try to make and look after our own fishtanks, at least not if our main goals are to enable as many people as possible to get into looking after, breeding and sharing fish, and if we want to be able to see a wide variety of fish. Wouldn’t it be better to have an ecosystem where we can all set our fishes free to swim together?
This way, everyone can access all of the riches of the deep and it’s a lot easier to get into fish research.
Of course, ecosystems can be dangerous places, with predators and diseases, and they can be fragile. You could also argue that what fishkeepers really want is the experience of nurturing their own fish, and the enjoyment of setting up and maintaining their own micro-infrastructure, and therefore fishtanks are the best solution. But there a limits to the applicability and relevance of any metaphor.
There are potential disadvantages to distributed infrastructures, and many of them relate to the additional complexity that they introduce into the access and identity management arrangements. Arranging access to services in one location can be hard enough, but authorization to use, for example, textual data in more that one repository might require passing of information between institutions. It is also the case that while there are reasonably well-established technologies and procedures and agreements for controlling access to online content, the authorization of web services is not such a well-established area. Furthermore, authorization to access online content cannot easily be passed on to authorize access to the computer processing power that is necessary to carry out an online textual analysis, if this is being provided by another centre in the distributed infrastructure. In summary, the fact that distributed services are reliant on cross-institutional agreements and arrangements adds an extra hurdle to be crossed to participate, as data provider or user, and a layer of complexity and additional layers of risk to the robustness of services.
Other potential disadvantages of distributed infrastructures include:
- Registering persistent identifiers with a shared service becomes desirable to sustain the interoperability of content and applications, thus adding another level of complexity to the curation of the data;
- Monitoring of usage is difficult, since operations are being carried out on remote servers not under the control of the repository;
- Monitoring of the availability of services is difficult – it might be possible to test the status of individual components but not a complex workflow;
- Although underlying interoperability is essential, there is no impetus towards consistency in user interfaces, and even a tendency towards heterogeneity, and therefore fragmentation of services is likely to be maintained or even made worse;
- Various further questions also remain (at least partially) unanswered in many cases, relates to where and how the computer processing is carried out, and how usage and services are monitored and logged.
We also need agreement at some level about our categories, formats and concepts. To get to the promised land, we need to agree on some standards. Linking datasets requires interoperability at the levels of the linguistic representations, annotations and metadata. Visualization of large datasets requires a reduction of variables, and deciding what is important and what is not. There is a tendency in the humanities for everyone to think that their way of looking at things and of categorizing things is unique. Annotations do sometimes embody the unique intellectual work of identification, categorization and interpretation of phenomena, and these are vital operations in the humanities, so it is not a surprise that this is problematic.
Another problem is that building infrastructure takes time and involves addressing complex and difficult administrative, legal, financial, political, technical barriers, often by making international agreements. So, usually, it’s easier to make ad hoc work-arounds. And building tools can be more attractive and rewarding. But actually, it’s a false opposition – enhanced infrastructure should help with tool development and deployment. An infrastructure providing a range of simple solutions for connecting together data and tools, deploying them as reliable services, managing authentication and authorization, licensing, access to computing power, monitoring availability, connection to virtual research environments, etc.
The mistake would be to try to build the perfect all-purpose tool, or to claim to provide services for end-users which solve all of the infrastructure issues. Or to put it another way, building the biggest and best fishtank in the world doesn’t solve the problem, because you can’t get all the fish in the world in there, allow everyone access to view every kind of configuration and interaction in there. But all too often this is what people try to do, rather than contributing a part of a wider, distributed system. Understandably people are impatient and our efforts and resources go into building new fishtanks, which can be fun to make, and which look good when people come to visit.