With side effects rife within Lisp's standard library, and with things like dynamic variables included with the language, I have a difficult time believing that it would be easy to "abstract out" side effects with some macros, while simultaneously providing a convenient, broad set of capabilities to program with. At that point, you might as well use macros to compile a whole new DSL within Lisp.
In order to quell the issue of "functional" having many definitions, the author chose one, and wrote about it. I don't see your alternative, which seems to be "accept that functional means something different and write about that", as one that is worthwhile to pursue.
People are not compilers. If you want to redefine a word that is already in your readers' vocabulary with a meaning ascribed you've kind of got to talk about why you're doing that and why your redefinition is worthwhile and the alternative, generally accepted meaning needs to be jettisoned. Or you risk unintentional impersonation of a crank, which would be a shame if you're actually correct.
You makd it look like he's trying to push a new definition, which is not true.
But if the definition is the author's own special definition, the conclusion may be true and still be completely uninteresting. The standard definition is the standard one for a reason - it turns out to be useful for having meaningful conversations about the issue.
But even if you accept the author's definition, then you run into:
> In fact it can be argued that lisp is one of the least functional languages ever created
For that to be true, "one of the least functional languages ever created" must be a really large set. That set must include, say, 95% of all languages ever created - which makes the claim not a very meaningful statement.
Like which? Pretty much every language has a ternary operator. Lisp's IF is not a function, by the way. A truly functional IF would have to wrap THEN and ELSE clause in lambdas so that all arguments can be evaluated before deciding which one of them to execute.
I am more used to thinking of it in terms of having function values (closures), but I think you could argue the lexical scoping that leads to SSA is all that's necessary once you have closures.
In any case, the vast majority of the lisp code I've read has been imperative and non-functional.
Come on. Returning a value from an (if ...) already makes it more functional than a lot of languages. Sure, it's not functional in the sense Haskell is functional, but you could certainly write all your code in a functional manner using it, abstracting out any side effects via macros. I would say it's fairly well-geared towards functional programming if one was determined to use it for such a thing.
That said, I wouldn't call lisp a "functional" language. I'd call it multi-purpose. It CAN be functional, it CAN be imperative, it CAN be object-oriented.
It's certainly not the least functional language ever created though.