User:Johncameronburns
Position Statements
Architecture Matters:
Digital Libraries in general are in transition. Too many are based on monolithic web-apps where the content, the indexing, the search technology, the presentation model, access control, and everything else is tightly entwined in an inflexible structure that cannot move with changing demands. And too many content providers find themselves having to repeatedly migrate to new platforms, and planning the next migration even before they have completed the last. The lesson being re-learned, as it was learned long ago in software engineering, is to separate the content, the rules, and the presentation.
Moreover, there is increasing realization that much value is derived from the inclusion of both formal, canonical content and grey content — user contributed content that complements, expands , clarifies or links the canonical content. This is especially true in mathematics, where perhaps the only people able to meaningfully contribute to the content are the practitioners themselves.
In the case of the DML a green-field site is being advocated, and the content base will not be so huge that the implementation must be optimized to the nth degree. We should therefore take pains to ensure that the content is exposed via standardized access methods, and that any GUIs provided use those methods. We should ensure that the content is machine accessible, and especially that the annotations and user commentary is as accessible as the primary content. Why? Because rigorous representation of the core content is likely to be provided in user annotation and open linked data, not in the originals.
Therefore we need to ensure that architecturally the content store supports a rich (and dynamic) representation repertoire, rich relationships, and rich attributes and access control.
Almost needless to say, the access primitives should be rich enough to support existing standards and general enough to support emerging ones, both humans and machine oriented.
Arguably, the GUI itself should be outsourced, since user experience (UX) is complex, messy, and arcane. Such a separation would also ensure that the integrity of the abstraction layers was not compromised.
Implementation issues
Having seen a range of implementations over the past several years, of the choices – in house, turnkey, contracted out, and generic platform – in-house seems to always be the most problematic. Generic platforms such as DSpace, EPrints, Fedora, and possibly iRods, seem to be the most cost effective, but with limited ability to customise, and limited back-office capabilities – and they still need hosting. Turnkey and custom platforms seem a mixed bag, but are always expensive to build and maintain. There are very positive experiences with turnkey systems, but there are also horror stories about delays, cost overruns, failure to meet specifications ( though often these are at least as much the the responsibility of the client as the provider).
It seems unlikely that WDML will even consider building its own platform, so its options would reduce to either a turnkey solution, a hosted collection inside an existing digital library, or a mildly customised open source platform. As mentioned, turnkey solutions would be expensive, and risky and likely will not be considered.
The Journal of the Society of Architectural Historians presents a good case study of how a rather challenging publication can be hosted whilst still retaining its own distinct identity.
DSpace and EPrints are well known and provide good options for a classical consumptive reading platform. The platform developed by the KB for the European Digital Library also looks worthy of investigation.
iRods is potentially very interesting. As a storage grid with executable policies it seems to promise much of what is needed to deal with the community curation elements that will be essential to anything more than a content silo. Again, it is worthy of investigation.
