Lambda the Ultimate

Blog post by Olin Shivers.

Excerpts with my editorial:

Newell and Simon's Turing award cited them for their work on “symbol systems.” Newell once told me that this was just names, and then he explained his understanding of names: “They provide distal access.” That is, a name is a local piece of data that stands for some other piece of data, which is presumably large and remote. You can now use that small, convenient, local datum instead of the large, remote thing for which it stands.

The treatment bothers me, because it doesn't distinguish very well between names for computers and names for people, which contrasts in the idealism I picked up from Edwards ("names are too valuable to be wasted on compilers"):

As far as Newell was concerned, this is the purpose served by names, or symbols, in all computational systems. Including the one in your head.

And afterwards, dives into the computational:

The BitTorrent example above is particularly interesting, since it comes with an unusual, distributed dereferencing mechanism. A BitTorrent name also has the additional exotic property of being a “true name,” in that it names one and only one piece of data.

He then visits the humanistic viewpoint:

Good name choices make it easy and natural to do the right thing—like expressive, well-chosen types, they lead you effortlessly to the terms you wanted to write. This is because names used by humans come with baggage. They fit into a larger framework. When you can get that framework right, and stick to it, then the names come easy—not just when they are minted, but when someone is trying to recall the name for some existing thing.

Queue a bunch of suspiciously OO names that take the form verb-noun or noun-verb. And then finally back to the computational viewpoint, where FP reigns supreme:

One final remark about names. We use names all our lives, every day, all day. So they seem obvious and not so mysterious. But names are subtle. There is a sense in which the λ calculus is nothing more, really, than a theory of: names. Go look at the fundamental, definitional rules of the λ calculus.

I'm of the opinion that the humanistic and computational nature of names are completely different, and find it weird that they are presented in such close quarters as to imply a strong connection between them.

Also, the way FP and OOP deal with names illuminates a lot about the paradigms. In OOP, objects have identities (intrinsic globally unique names that allow for easy aliasing) with interfaces that are named according to natural metaphors; FP is rather more oriented to symbolic reasoning where values (in pure FP at least) are anonymous and defined strictly by their structure, functions are named according to transformations on this data (map, reduce, select, etc...).