Something that stuck with me in the dialogue so far which seems important (and I don't recall who introduced it), is the idea of a compiler or an interpreter that simply refuses to compile syntactically malformed code. This really foregrounds the implicit difference between "code" (as in computer programming languages) and "language" (as in "natural" human languages uttered/written in the world). Theoretically, programming code can have all the robust, affective wiggle room of human languages -- in other words, it can have the ability to be "misread."
BUT... not all "misreadings" are equal.
There are the affective misreadings of uttered human languages in the world that allow for qualitative differences in the ways in which such language is interpreted and responded to. For instance (as has been mentioned), "misread" or "misinterpreted" instructions to a dancer don't necessarily binarily and fully inhibit the dancer from proceeding to move at all. Something can always be performed in the world as a result of these kinds of utterances, however "misinterpreted." Actually, a billion things can be performed. And each different thing performed qualitatively differs from the other things that could have been performed. Because human bodies are analog meat matter (at least in some non-incidental sense).
In these real world contexts and systems, language acts as a kind of affective, analog force. Not simply because the syntax of human language systems are infinitely open, undecidable, and complex. That's only part of the reason. Because even with very simple instructional language systems, an almost infinite number of interpretations and resultant behaviors are possible. ASSUMING THATů the interpreters are themselves complex (and humans are), and ASSUMING THATů the interpreters are embedded in overlappingly complex contexts (within an almost infinite array of telescopingly complex contexts). And humans are always already embedded and implicit within such contexts (just sticking with the example of human dancers -- contexts of etiquette, art historical contexts of performance and dance, contexts of skilled/deskilled athletic performance, gendered body movement contexts, modestly contexts, etc.) Heck, moving beyond the dancers and their analog bodies for a moment, just shift to any sentence uttered in any coffee house conversation between two people. That sentence can be "interpreted" and "compiled" by its interpreter (the listener) in an almost infinite number of ways. It's not that every sentence is completely arbitrary and open to any meaning. The history of natural human language use constrains and colors possible present interpretations. It's just that even within these historically contingent/constrained semiotic boundaries (of an utterance even as simple as "i love salt"), the shades of subtle differences in interpreted meaning can be almost infinitely fine-grained and nuanced (depending on telescoping contexts, interpretations of tone, bodily affect, prior human relationships between speakers, etc.)
Compare these types of human language interpretations to the compiler in a computer "misinterpreting" code and returning some sort of error. I realize not all compiler errors are equal. A buffer overflow error could be some run-time error where no error message was ever returned. It could be a syntax error, or a logic error, or whatever. But (depending on the programming language and the compiler), there are still only so many ways such code can be "misinterpreted." In the case of esolangs (esoteric programming languages), these languages are meant to remove such binary safegaurds and allow the system to corrupt, because these art languages are more interested in qualitative flavors of misinterpretation than they are in running a program that functions "properly." In the case of "glitch art," the artist is looking for the liminal zone in which failure can fail in a qualitatively infinite number of ways. But (as observed by Daniel Temkin), even such "glitch art" is not a fail to the computer. The computer is just running as it is meant to run, according to the way it was meant to interpret the instructions ("databent" to us, but not to it) that it was given. And we humans "interpret" the resultant "glitched" visuals as failure. But they are not failure "to" the machine, because when the machine actually fails/misreads according to the way in which the machine it set up to "interpret" failure, it simply returns a broken gif icon or refuses to open the corrupted file. A binary 0.
So how to get the machine back to the affective, qualitative variability of the (weird) world; how to cause the machine to "actually" misread (rather than merely return a symbolic re-presentation of the misreading event)? Well, you could always code in assembler, or in hex, or in binary. Surely those root-level code languages would free us from the arbitrary, human-imposed digital binariness of human-built interpreters, and allow us to once again return to the analog, affective, qualitative continuum at the heart of matter in the actual world. But no, the controlling [meta-]context of a digital system is set up to fend off (or, more precisely, round off) any traces or remainders of affective difference. Yes, there is a physical location in the material of a silicon chip that is either charged or not charged. But it's probably not so simple in the physical world. I myself don't know the physics of sand, but I imagine with instruments sensetive enough, for a binary 0, the sand would probably be charged 0.00001 % or something. That affective physical "difference" just gets shaved off and ignored. It doesn't get "read" or "misread" in any nuanced, interesting way. This shaving is at the heart of digital computing (although not at the heart of analog computing).
So a vacuum tube slightly warmed up but not fully warmed up = 0 in a digital system. But a vacuum tube slightly warmed up in a marsall tube amplifier returns a qualitatively different type of distortion than a vacuum tube a bit more warmed up than a vacuum tube a bit more warmed up, etc. And a mostly punched hole in a punch card = 1 in a digital system. But a slightly punched hole in a punch card that is "interpreted" and "compiled" by a bunch of humans in Florida within the massively entangled cultural contexts of a US presidential election is not a 1 or a 0, but the inordinately problematic "hanging chad." Because, according to the logic of such binary political enframings, "if you're not for us, you're against us." Any in-between qualitative differences are not "misinterpreted" or "misread." Such differences are not even "ignored" (because "ignored" implies a kind of calculated, triaged, willful non-recognition on a case by case basis). Such differences simply fail to register anywhere within the artifically created system. They are shaved off by the system prior to even being considered as differences by and within the system. Such differences are turned away, every time, without exception, at the very gates of the system.
But digital systems are not the affective world (although they exists within the affective world). Likewise, language does indeed function within digital systems, but its functioning within those systems is just a kind of stilted subset of the way that it functions in the affective world. The way language functions within a digital system is by no means an "artificial mirror" of the way it functions in the affective world. Indeed, how could it be, since digital systems are run on, within, and are encompassed by the larger affective world.