11:35 <msiism> Up until now, I used to have a "GLOBAL CONSTANTS" section at the top of my Racket programs that was divided into two subsections: "Literal" and "Computed".

11:36 <msiism> Now, I've learned that none of this makes good sense.

11:37 <msiism> But then, not all of the global parameters I define at the beginning of a program are "of the same type" (in my mental model).

11:38 <msiism> Some of them are not meant to ever be chnaged during execution, while others are.

11:40 <msiism> Now I'm looking for the right term to make that clear in the comments. Maybe there are some well-established terms already.

11:40 <msiism> Here's an example of what I used to do: https://git.sr.ht/~msi/rpack/tree/current/item/rpack.rkt#L34

11:42 <msiism> I'd go with "GLOBAL PARAMETERS" for the top-level comment (heading). Not sure what to call the sub-sections, though.

11:59 <winny> msiism: This might be of interest to you https://docs.racket-lang.org/guide/parameterize.html

12:05 <tonyg> Mind candy, for those of you with an afternoon to spare for thinking about real numbers and their representation in computers: https://posithub.org/docs/Posits4.pdf

12:09 <msiism> winny: I've looked at this a bit already. But I'm still not sure about the actual benefit over setting things up with `define`.

12:09 <msiism> I mean, as far as I've read, using a Racket parameter does not guard against a value being reset.

12:10 <msiism> Err.. I mean a variable being reset.

12:14 <winny> I'm not sure about the tradeoffs. The parameterize technique seems to make it easier to write tests (because the environment is scoped to the tests [using continuations]). I'm not quite sure if it's possible to use a binding (using "bare" `define`) across racket module boundaries (between files for example)

12:14 <winny> s/to use/to update/

12:16 <winny> It's certainly a pretty unique idiomatic technique to racket.... Envisioning there should be an article comparing and contrasting various techniques to track global configuration

12:17 <msiism> As far as I've seen, you can't `set!` a parameter initialized with `define` that comes from a module (which is good, IMHO).

12:23 <msiism> Just re-checked HTDP and it introduces `define` as a way to define "global constants".

12:23 <msiism> But that's in BSL.

12:26 <msiism> I mean, my idea of a constant is that a name is assigned to a value (or vice versa) and that that assignment cannot be altered after the initial definition.

12:30 <msiism> Well, the constant is actually the value itself.

16:22 <samth> tonyg: note this critique: https://hal.inria.fr/hal-01959581v2 my sense is that posits are potentially interesting in some places but not really a big change

16:35 <tonyg> samth: thanks for that link! very interesting