> For context, that was dev-luatex list not the main luatex list at tug
> I'd have thought that your example isn't convincing, and that parshape
> should always follow the writing direction. Even if the shape is a map
> of the Netherlands if writing left to right or right to left, specifying
> a sequence of starting points and line lengths would seem reasonable,
> although of course the actual numbers would be different depending on
> whether you are specifying offsets from the left or right.
Nothing prevents us from having two versions, of course, but
I was describing how \parshape is being used, and the fact
very likely both uses are going to coexist in the future. I
still think the default behavior should depend on the current
direction, but we must not forgot this fact.
> even if luatex's primitive \parshape doesn't get a flag, you could
> define it in lua,
I've even managed to write a "converter" (just a "promising" draft) for
the case n = 1 in pure TeX (with a few \afterassignment's). And the
new tokenizer is of course another solution (thank you for the code).
But for backwards compatibility the issue of the two functions
still remain, and I'm not sure how to handle the possible conflicts.