For blocks there was a special form like TO repeat :count [block :expr], and similar to to arity the function definition would effect the parsing; lists in the :expr position would actually be passed in as anonymous functions. Again, if you look up all the function definitions up-front this would work much better.

In a Tcl translator that I wrote it did the translation lazily (except in cases where it knew that a block was called for, like proc), and associated the translation with the string. This seems like a reasonable strategy for Logo as well. You could even compile all lists present in the source code, and simply ignore errors when they occur, and only actually signal the error at runtime if the person does use the list as a block.

Another strategy, if you don’t mind diverging significantly from standard Logo (and NetLogo is already very divergent from the brief example you link to), is to do something more like Smalltalk, where code blocks have a convenient syntax. I believe Smalltalk actually took inspiration from Logo in this regard, using [] as a kind of lambda. You would need to get another syntax for literal lists then, but literal lists aren’t that interesting, and most other uses of lists in Logo tend to be substitutes for strings, e.g., print [hello world!], and you’ve already got syntax for full strings in NetLogo instead of only words.

NetLogo is an unusual Logo; much of what you say about Logo in this post doesn’t apply to NetLogo. The differences are summarized in our FAQ at http://ccl.northwestern.edu/netlogo/docs/faq.html#logodiffs . In particular, NetLogo is lexically scoped and control structures are special forms. This brings us closer to Scheme, though at present we don’t have a macro facility or any other way for users to define their own special forms. These differences are there partly for design reasons and partly to facilitate efficient implementation.

NetLogo is partially compiled, partially interpreted. It’s reasonably fast — much faster than a naive, totally-interpreted implementation would be. All parsing is done at compile time. Our compiler generates Java byte code from snippets of user code, but it doesn’t handle the whole language yet, so we still have an interpreter to stitch those compiled snippets together. Over time, we plan to get closer and closer to full compilation.

There is so much to say. I wish I had paid more attention.

This is from a guy who saw his first computer in 83 and then bought an Apple IIgs because color monitors were cool in 86.

Now I am a damn .NET developer because that’s where the jobs are in my area.