Alexis King

In a perfect world, I think `ReadableHashTable` would be the supertype (and we could just call it `HashTable`), and `ImmutableHashTable` and `MutableHashTable` would be subtypes. You'd get flat contracts with...

Please read the documentation for those primops before using them—using them with arbitrary `IO` code is not safe.

> I believe `bracket` is well-behaved wrt `control0#` No, it is not. A version of `bracket` that was designed to support delimited continuations would act like Scheme’s `dynamic-wind`, which the...

> So when `action` is executed, it'll just not work at all, because after the first branch finishes, the effect is no longer there. Yes, this strategy just doesn’t work...

I personally like the magic, since it doesn't have any downsides for those aware of it, but I also recognize the potential confusion, so `and*` and `or*` would still be...

From a certain perspective, I agree. It would be absolutely possible to add magic in places where magic does not belong, to the point of making the language incomprehensible to...

@mflatt Sure. For `racket/base` specifically, the difference isn’t super relevant. However, for other `#lang`s, it might be. The idea is that `#%namespaced` is much “lower level” than something like `local-require`....

That all makes sense to me. `#%namespaced` is a hack. It’s a hack that solves a problem, but it’s still a hack. If you think that there’s even a *possibility*...

My understanding is that we are waiting for the Racket rewrite of the macroexpander to become *the* macroexpander, which is currently waiting on the Chez rewrite, so this has been...

> I find it confusing that a metalanguage hooks into only the reader but would expect to behave as both a reader (that is, converting bytes from an input stream...