Site Architecture
In his keynote speech to the 2002 Dublin Core Metadata Initiative (DCMI) conference in Florence, Italy, DCMI Executive Director Stuart Weibel likened the architecture of the web to that facing Brunelleschi in designing his famous Duomo (right): at the time that Brunelleschi designed his dome, building technology was not yet available to build it. Despite that, building work started on the massive cathedral and, when the time came to start the dome, technology had caught up.Stuart used this metaphor to address some of the troubling concerns around information architecture: sketch out your desired architecture, whether the technology is available or not.
In designing this modest site, I have, in all due modesty, attempted to apply certain key architectural principles, and avoid the pitfalls that I have seen so many times with web sites built with haste, with proprietary tools that give the user little control over the design, or with little thought for user navigation or scalability.
Firstly, no "hard-wired" hyperlinks: every piece of writing is considered as a "logical object" and it is that object's unique identifier that is "exposed" to the end user, not a specific filename. For example, this page has identifier 0019, and always will have. If the text is modified and updated, as long as the main semantic intent of the contents remains the same, it will always carry the same identifier. If it changes, then it is given a new identifier.
What is delivered to the user is rather a "representation". By default, that will be the most recent version of the English-language, html format of the content, but a file, once created, will not be modified. The logical object will instead point to a different file if content is updated (with links back to other representations being available, if required).
Secondly, associate metadata as a priority to the logical object and not the representation. As representations come and go, only metadata that offers additional value should be associated with specific representations.
Thirdly, keep options open regarding knowledge management and semantic web technologies, rather than hard-wiring information into the objects. RDF, I believe, encourages a bad practice of detailing associations between resources within the resources themselves. If the metadata are clear and comprehensive for all objects, it should be possible to construct an ontology of the site, whether using RDF, Topic Maps or OWL. Furthermore, the anatomy of the "hyperlink" - a "triple" of an assertion ("there is a link between"), and two resource identifiers ("source" and "target") - has blinded the RDF community to the fact that associations can often be far more complex than triples, and involve many parameters and axes (for example, "this person, at this time and location in this context, is associated in the following ways with that object, at that time, location and conditions...").
Fourthly, keep representation (the manner that I have structured the information on each page) separate from presentation. In this, the W3C specifications excell: I use a limited subset of Cascading Style Sheets level 1 (CSS1), to manage all aspects of the site's presentation. In this way, I have complete control over the visual and spatial design of the site, without having to worry about re-writing any of the site's pages.