http://www.erlang-exchange.com

With over 30 speakers, including Erlang Creator Joe Armstrong, Dennis Byrne (JInterface creator) and other experts, presenting talks, practical workshops and leading discussions on Erlang, concurrent programming and functional programming, this 2 day event in London will be the ultimate platform to learn, network and exchange ideas!

Tickets go at GBP 300, but if you register on or before May 1st, they are just 250. Tickets are limited to just 200 people and, particularly the last 2 weeks, tickets are going very very fast, so if you fancy coming, get your skates on!

More information and registration: http://www.erlang-exchange.com

Obviously any classification of languages is going to be simplification of the reality. All sorts of other classifications are also possible. I think it usefully illustrates a point for this article nonetheless.

Not so. It’s trivially easy to show that programming languages can’t be sorted into four neat little piles. The piles overlap. Like anthills built too close together, they share both outer surface and inner features. Prolog has functional features. CAML and Java have procedural features. Common LISP has a bit of everything. In fact, I believe your point is that functional features can be injected into the language of your choice.

Not only do the categories dissolve at the edges, they aren’t sufficiently bifurcated. Following Colmerauer, Prolog is a special case of constraint programming. To “fix” the diagram, CP should be pushed to the top and Prolog down a level, with its peers added. There are “pure” OOP languages like Smalltalk and impure ones like Java. This may be worthy of an additional level.

But this node-splitting produces increasingly disturbing results. For example, suppose instead of splitting on the attributes “pure” and “impure” one chose to split OOP on “(not) suitable for concurrent programs”. This would surely put Java in a separate pile from C++ and both rather far from the pile that contains Erlang. If instead one split OOP on “(non) contractual”, Eiffel would be king of its own hill while all the others would find themselves roomies.

The notion that programming languages can be sorted into a simple taxonomy is one of the persistent errors of computer science, as is the notion that issues with one branch of the tree can be resolved (or at least greatly improved) by an infusion of ideas from another branch. It is too bad that the 80’s fumblings of the Microelectronics Computer Consortium (MCC) are not more widely publicized. Then everyone would know that you can’t make a great leap forward by crossing functional and logic (FOOL) or all three of the not discredited paradigms (FLOOP). Instead, you get a mess.

Any useful sort on programming languages must be multi-dimensional, in terms of the attributes that give each language a reason to exist. Phrasing the debate in terms of four tired categories is a habit we all must break.

I didn’t realise there were multiple implementation. Thanks for the clarification. It would be interesting to compare the two implementations for performance and number of lines of code.

There is an implementation of Perl 6 being written in Haskell called Pugs, and there is also an implementation of Perl 6 being written in Parrot called Rakudo. For more information about Rakudo, see http://www.rakudo.org/

y = f(x) * f(x);

You really mean:

y = f(x) + f(x);

Else the other stuff underneath it isn’t quite true. Either than or you mean:

y = z * z;

(I love scheme, I really do…

There was a mistake. I have fixed it now. Thanks.