Actually, I'm not so sure about this split: M(X) is only defined inside this single paragraph at the moment. With your change, we would leak this expository device into the whole-subclause scope, if you see what I mean.

This is recent CWG text and I concur with @tkoeppe's view that M(X) is local expository device that we should not spread more.

I see what you mean, but the current formatting is ... let's just say it leaves a lot to be desired.

Why do we say "(defined below)" if it's right there in the next sentence? We already say "defined as follows" in that next sentence, how many times do we need a teaser trailer for that definition?

Why do we have that "M(X) is defined as follows" floating in the middle of the list, unattached to any single bullet?

Why is (8.8) a continuation of the same list, if it only applies to (8.7)? Wouldn't the bullets defining M(X) it be better as a nested list, e.g. (8.7.1), (8.7.2) and so on? I know we can't do that with our current itemize env, but here's what it looks like using enumerate instead (rebased on master, so p8 is now p9):

(That doesn't change the fact it's excessively long, but I think it improves the structure)

Correction, the example doesn't need to be nested under bullet 7, it applies to the whole paragraph:

The example could even be its own paragraph with a new \pnum:

@jwakely, the "enumerate" stuff is a regression, not an improvement. I'm ok with a nested "itemize" to define M(S). And when you do that, the "(defined below)" should really go. Asking @zygoloid for an opinion.

I like the last version, with the M(X) part of the same paragraph where it's used, and the example separate.

I'd phrase the relevant item as "has no element of the set M(S) of types as a base class, where M(X) is defined as follows:".

Or even:

has no element of the set M(S) of types as a base class, where for any type X, the set M(X) is defined as follows. [Note: M(X) is the set ... zero offset in X -- end note.]".

I'm not suggesting we actually use enumerate, obviously we don't want numbered bullets. When I tried a nested itemize it didn't number the nested bullets properly so I just used enumerate for argument's sake. But I see what I did wrong now (I still had the new \pnum after bullet (8.7)). Here's what I meant , with nested list, example in a new paragraph, and no "(defined below)":

We could even join the two sentences in bullet 7:

Or:

@tkoeppe my edits crossed with your comments, so I'm not sure which "last version" you mean now :-)

I won't edit it again.

I mean the last screenshot in #2066 (comment). But what you have now is equivalent for the purpose of the stuff I was talking about.

This is a drive-by, but we really should say what X is in M(X): "For a type X, M(X) is defined as follows..."

@jensmaurer: I just suggested that above :-)

+1 to something in the direction of @tkoeppe's suggestion:

has no element of the set M(S) of types as a base class, where for any type X, the set M(X) is defined as follows. [Note: M(X) is the set ... zero offset in X -- end note.]".

I've just pushed a new commit that results in:

Oh, can we fold the footnote into the note?

The footnote is a somewhat imprecise rationale for the rule, so I think it would need quite a lot of massaging if we wanted to promote it to a note. Maybe we should consider removing it. But I think this PR is already a significant improvement, so I think we should merge it as-is and leave the footnote for future cleanup.