Hi Peter,

> I don't know - those SRFIs seem overly dynamic, using things like
> predicates and disjoin/union.  The CHICKEN type system is a lot simpler.

Predicates can easily be converted into more exact types whenever
necessary. It’s already done in SRFI 253 implementations for CHICKEN and
Kawa (the only implementations that provide strict types that I can get
my hands on, besides Gambit.)

> For most SRFIs packaged as CHICKEN eggs we already define some types,
> but it's using the CHICKEN "external types" file, which is a purely
> static way of declaring types (a bit like declare-checked, but with a
> predefined set of types).

Yes, that’s a good practice! I’ll look into these types (pun) of files,
to see if one can steal it.

> For more info on the type declarations themselves, see this:
> https://wiki.call-cc.org/man/6/Types

I know of these, and they are already used in SRFI 253/273 macros. To
much joy.

>> Is there a need for such a thing?
>
> It would be even better if new SRFIs came with type declarations in the
> reference implementation.

Yes! And that’s why I suggest SRFI 253/273: it’s a widely-implemented
and portable SRFI family. And the “dynamic” nature of them you talked
about helps here, as SRFI implementors don’t need to remember syntax of
type systems for any and every implementation. Which is a burden, with
at least three implementations shipping their own incompatible syntaxes
and type categories!

Thanks,
--
Artyom Bologov
https://aartaka.me