In 1994, Stewart Brand wrote a book called How Buildings Learn, which explored how the built environment changes over time as people use it. In it he brought to light a concept known as shearing layers, which, like the different hands of a clock, map to the rates of change of different aspects of a building and how those aspects move largely independently of one another. Three years later he published a six-part documentary, the last of which covers the subject in depth:
Temporality is an exceptionally useful axis along which to organize. The speed at which a given aspect tends to change dictates how much effort we put into it, how much deliberation and political alignment is needed, how much we invest in durable materials, and how much of the aspect is subject to forces outside our control. The idea that different aspects of a system change at different time-scales that shear against one another is also enormously valuable—it means that in changing one, we have to account for its effect on the others. This is just as applicable to the Web as it is to buildings. In fact, it appears to map perfectly.
It turns out that I'm not the first to apply this idea to the Web. Peter Morville, for instance, did ten years earlier than I. I'm sure many others have as well. I want to tell you why I believe it fits.
What Brand means by site is a definite region in space where human activity occurs. A site may be occupied by many buildings and other fixtures that facilitate this activity over an arbitrarily long period of time—in his mind, millennia. As such, it would be awfully parochial to draw a direct parallel between this concept and a mere Web site.
A geographical site exists very plainly in physical space, but the kind of site we're concerned about occupies a much more ephemeral, conceptual and representational space. Both produce the same kinds of important social, cultural, psychological and semiotic environments, but while one is reinforced with stone, wood, concrete, glass and steel, the other can be vapourized on a whim or common accident.
I believe it is absolutely essential that when considering the site, that it isn't really just what we normally call the site, but its entire neighbourhood of other sites. When people contribute to the Web, they do it because they want their message heard. When other people link to that message, especially from another site, it is a vote endorsing its attention (irrespective of whether that attention is praise or criticism).
Taking this greater concept of site into account, we can frame a concrete set of considerations for the environment surrounding what we normally dub a site.
When Brand talks about structure, he's referring to the constraints and affordances of the materials used as well as the properties of the forms that can be composed out of them.
For example: The Web exhibits in great abundance one affordance that nearly every other medium either does poorly or not at all: intension. It pertains to understanding and connecting concepts together in terms of their content. It is the complement to extension, which pertains to understanding concepts in terms of the categories they belong to. Intension is about free association, extension is about classification. What information architects tend to call structure is really just different forms of extension, from hierarchies, to facets, to search results. But as people we use both extension and intension to understand the world around us.
We can understand these properties and put them to work for us, or we can ignore them and risk persistent conflict.
The skin is whatever comes into contact with the outside world. This isn't just the visual accoutrements intended for human consumption, but the interface for the growing proportion of non-human consumers as well.
The skin has semiotic value as well as aesthetic. Its features mark points of interest and enable people to orient themselves. It is perhaps more important to exist as an unambiguously recognizable idiom than to invoke a pretty metaphor. Machines can be more easily confused, requiring an even stricter adherence to a well-defined set of semantics, and any sudden movements are abhored by both machines and people.
Therefore: Make the skin reflect the structure—entrances and exits, points of transition and inflection, public and private space—but the structure need not dictate more than it must. Decouple the skin from every other consideration, so that it can be dealt with on its own, and so other elements may not unduly upset its integrity.
Services consist of the systems that actually make things go. They are what enable people to get things done. In a building, they provide people with power and water, make the rooms hotter or cooler, or move people throughout the structure. They are analogous to applications.
We can say a few things about services: how you get a job done is never as important as the fact that you get a job done. Likewise, how you do a certain job tomorrow may not be the same as how you do it today. In other words, technologies, platforms, products, languages, styles and maintainers with their pet preferences all come and go. The services plan should account for that.
Moreover, services are inherently heterogeneous, with parts being removed as new, unanticipated ones are added. Trying to coerce that behaviour into a monolith is hard, costly, and underperforms if it performs at all.
The space plan is what we would normally refer to as structure. It comprises the sections, the site map, facets, search, visual hierarchy, content positioning, et cetera.
In a building, you adapt the space plan to the human activities that happen most frequently. On the Web, there is little stopping us technically from tracking precisely what activities happen most frequently and adapting ad hoc and doing that for each visitor or discernible group thereof—if only it were that easy.
Changes to the space plan of a building are heavily moderated by the material cost of modifying non-structural elements, but beyond that, what to do tends to be easy for everybody to understand and organize around. Changes to the space plan of a website or any other software system, while they carry a negligible material cost, typically involve coordinating several constituencies with diverging interests, varying levels of understanding and vast disparities in both the domains and ranges of their respective executive power. More often than not, there isn't a single person who can control the entire space plan of a website, though this is a problem that can easily be remedied if we consider it at the outset.
Stuff is what pertains to what actually goes on inside a space. We might also recognize the concept as content. It's the reason why we're here. So let's get everything else out of its way.