Am Fr., 31. Juli 2020 um 20:49 Uhr schrieb Wolfgang Corcoran-Mathe
<xxxxxx@sigwinch.xyz>:

> OK. If John concurs, we can remove the constraint on indexers and
> cherry-pick the earlier composed-indexer version of the sample
> implementation.

At least for range-split-at we need composed indexers anyway. And for
subrange probably as well.

* And in view of this, exposing the lower bound to the user doesn't
look so natural anymore. It could as well become part of the closure
(or abstract) indexer, couldn't it?

* Moreover, I would propose to add

(range-join RANGE ...)

with the obvious definition.

* Ranges do not have to deliver a monotonous sequence. So the overall
name "range" and the predicate "range-includes?" are somewhat
misleading names. Most of the API is about (algorithmic) finite
sequences.

Marc