Lambda the Ultimate

From this web site: "I am using facilities that IE has which other systems don't, to maximise the amount of information I can get on a screen."

Which on my computer (Mac OS X, Safari browser) means zero information.

I've updated the taxonomy keyset page, so it's now appropriate to the dataset.

To people who don't want to use the same browser I am currently using, sorry. How about going to a cybercafe where they have a suitable browser, grabbing a cup of coffee, and looking at it there instead?

The material has come first for me, and when I have finished the rest of the data-input (18,000 more reference, 1,700 more languages) I'll fix it up some. (and the pages are more or less equally buggy cross-platform-wise)

I'm a total taxonomic junkie (my mom's a mycologist, I've been collecting programming language curiousa since I was in high school), but I'm also a web browser collector, and of the four browsers I can use, I can't even get around the site in any of them. I get not wanting to get wrapped up in any flamewars, but I have everything except Windows... it's a little frustrating, you know? Especially when what I can see looks so tantalizing...

I have prepared an alternative entry point at http://hopl.murdoch.edu.au/backdoor.html

I have added a "click here for simple version" reference at the top of every language display page to enable you to see the details in one long page.

I have made a simplified version of the examples to show on their own separate page.

This should enable the complainants to see the information without frustration.

The CLIST page doesn't include the fact that it's an IBM language. It does say TSO, but if you search the notes for IBM, CLIST does not appear on the result page.

I am curious where the Oz language would fit in your
taxonomy. Should the taxonomy be a graph instead
of a tree? Oz was designed to bring together on equal
footing concepts which are usually treated on unequal
footing.

The Oz design exposes interesting new axes for the
taxonomy. One axis is different approaches to
concurrent programming (declarative, message-passing,
shared-state). Another axis is different approaches to
data abstraction (object style, ADT style). A third axis
is the semantics used for reasoning, which depends on
which concepts you choose to program with
(denotational, operational, axiomatic, logical). All the
members of these axes are well-represented in Oz.

For these reasons, Oz was a good choice for our new
programming textbook. For a quick overview of Oz
check out the talk on Teaching Programming.

Ok, after thinking for a bit I'll propose a partial
answer to my own question. In my view, an
important part of the taxonomy should be based
on programming concepts (in a precise sense
that is made clear in my textbook). Languages can
then be classified as to how they treat the concepts.
Usually, languages will make it easy to use certain
subsets of concepts, for example, because they
simplify mathematical analysis (like functional
languages) or because they simplify building
modular programs (like object-oriented languages).
A language that emphasizes a certain subset of
concepts over other subsets can be given a label,
such as "lazy functional language". Not all possible
subsets are useful; experience shows that some
subsets are more useful than others. And the
subsets are overlapping of course. When you
finally pick up the pieces, this gives a lattice
structure where each node is a useful subset, and
languages live on this lattice.

Well, monothetic classification schemes serve properly to classify, rather than polythetic ones which map instead. If we want to group languages polythetically (which we might) then we need to have an acceptable set of values which apply equally to all (not something monothetic systems need worry about).

Then, as we know from taxonomy and explanation theory, we can't deduce the structures used bootstrap-wise. Declaration of an intent to merge approaches may be genuine, but that does not necessarily mean that a language created is a genuine hybrid. (They do exist - the classic ones are the merges between discrete and continuous simulation systems. However, they always have a comfort zone for one or the other). The question is one of applicability, or of prime appositeness, rather than absolute ordination.

And I worry about the mixing up of values, as category errors. It may (again) be the case that distinctions between syntax, structure, level etc are separable, but it has to be argued for, not stipulated or presumed.

Some of you have suggested a mathematical/cladistic approach - again, it has to be argued that the saime rules for distribution of features in the biologcal realm apply to the things in the cultural. This is an ongoing debate in prehistory of history of ideas, but has to be argued for.

I haven't finished writing up the justification for the taxonomy yet (a bit surprised by the overwhelming response) but I am trying to approach the task after the fashion of sociocultural taxonomies, by considering context, telos and inheritence, as well as documentation, form and intention. Also try to follow the art-historical way of investigation, proceeding with these as art-cultural objects.

Main reference for systematic cultural taxonomy is Dunnell, R. C. "Systematics in Prehistory" 1971 Free Press

So in summary, I think a graph or a network only works for influence chains, rather than typology - there is certainly no natural (prebioengineering anyway) equivalent. And a lattice doesn't work for locating (and would still have to be weighted mathematically)