ThoughtStorms Wiki :: ChristopherAlexander

Since christopher alexander is back in The Discourse™ i'm thinking about the main (deliberate? disingenuous?) criticism/misinterpretation of his work, which is that he is some kind of nostalgic neotraditionalist who wants to take the built environment back to the oldtimey days…

…this overlooks the fact that his doctoral thesis was on the topological decomposition of design problems using a method we would now call "min-cut", something only achievable using a computer…

…software people tend to latch onto the patterns (the single best-selling book out of oxford university press) and despite taking a decade to develop, they were only really an ultimately inadequate stopgap solution, and he even said as much himself

…really though you can trace an arc from the mathematical work in his 20s and 30s through the patterns and into the nature of order (fifteen properties/structure-preserving transformations, "fundamental differentiating process" &c)…

notes on the synthesis of form (a thesis on min-cut graph partitioning) argued that every (design) problem has an inherent structure, and there are objectively better and worse ways to subdivide the structure (recursively) into subproblems that are ultimately tractable

the pattern stuff can therefore be viewed as a set of connected subproblems that have been pre-disarticulated along relatively min-cut boundaries

(also worth noting that the book was called A pattern language not THE pattern language; it was meant to be a prototype for other pattern languages people would either adapt or make from scratch; why it's called a "language" is it's a set of shared concepts for discussing)

anyway it's a shame software people latched onto patterns and didn't really look much past them:

• the later CES books have a lot of practical things to say about procurement, contracting, project management
• the nature of order describes a "zero-cut" incremental process
• notes on the synthesis of form tried to answer the practical questionof "how do i break a complex problem down into something manageable while destroying as little information as possible?"
• nature of order describes how to do that without resorting to phd-level math

…those who would paint alexander as a nostalgic traditionalist fail to take into account that what he's ultimately indicting is the building process invented in the 20th century that puts an immutable drawing as the authoritative legally-binding reference document…

…in other words the drawings dictate the benchmark for performance and nonperformance: if you don't build to the drawings, you don't get paid—rather you get sued—it doesn't matter what new information comes along; it doesn't matter how insane or stupid the drawings are

i guess my question to software people is: does this sound at all familiar?

the first alexander book i read was the dissertation (even though i had had a copy of pattern language for most of a decade); at the time (2007-08) my my focus was on the performance risk of software projects (ie cost overrun/benefit shortfall, ie time estimation)

i suppose i should finish this thought: i saw Notes on the Synthesis of Form as a potential way to disarticulate complex (one might say, "wicked") development problems in the least damaging way, such that the elementary parts could be weighed and summed up

to make a long story short, what i discovered was that if you set upon this task, at least when you are working with a medium made of pure thought, once you are finished there is very little work left to do

in other words it is totally useless for mitigating performance risk

alexander's later work (nature of order) does not refute his earlier work (notes on the synthesis of form); rather it sublimates it