Why is it that most esolangs seem to be either 'fungeoids or Turing tarpits? Not that there's anything wrong with either class, I'm just wondering why are those two the most popular ones. Now, I do realize there are other kinds of esolangs, but they seem to be the minority. I just happen to personally like languages like INTERCAL, ILLGOL, Bear Food, Magenta, Homespring (which is/was ingeniously insane even if it is tarpit-ish) and the like. Most of them give the outward appearance of sanity, but when you've read past the first few lines of the specs and sources, it begins to dawn on you that you've stepped into the Twilight Zone.

Anyhow, that's my rant. Have fun, kids.

Because languages you don't have to write a parser for are easier to implement.

I've been thinking about this lately and it seems that most esoteric languages are more like "esoteric bytecodes." Sometimes literally: a BASIC compiler exists that targets Brainfuck; the author of Lazy K writes that "Lazy K... is better considered as bytecode for an unorthodox virtual machine."

I too would like to see more esoteric languages, but they are a lot of work to write an interpreter or compiler for. Several people have seen my spec for a C-like language that operates only on bits and arrays of bits. I haven't gotten around, though, to figuring out lex and/or yacc so I can parse it.

I think graue pretty much nailed it down. Parsing is a lot of work. Heck, I thought that simply parsing integers was perhaps not worth the hazzle when I designed Kipple (though when I actually implemented it it was rather trivial).

Excellent point. It's a shame all paraser/compiler generators are such a pain in the arse to use. Even the ones made for Ruby ryacc/rlex suffer from the fact that the original developer is Japanese, and so the documentation reads like a Babelfish translation.

YACC and Lex... well, don't get me started on those :)

Well, personally I don't like much about long instructions. I prefer one-character instructions. Anyways, I wouldn't call that kind of languages 'esoteric bytecodes', I can't see a reason why languages with one-charater instructions couldn't be called languages.
Even if making interpreters is hard, that doesn't answer the question why there are so many fungeoids or Turing-tarpits. I assume making a language Turing-tarpit is somekind of honour question. I like minimal languages, and they don't need to be Turing-tarpits. It's fun to have something else for a change, like for example Sceql. I think I'll want to develop some small low-level non-Turing-complete language next.. :) There are many characters in ASCII to use and many computation models to try out, et cetera.

I was referring to the fact that Brainfuck, Sceql, et al. are not really linguistic. When you write in a human language, you have words and punctuation that you assemble into sentences, and it's easy to see the analogy between that and, say, Ruby, C, Haskell, or INTERCAL. With the bytecode-like "languages", there's nothing comparable.