Ranges with length that cannot be named

[berni]

On Wednesday, 25 September 2019 at 17:03:07 UTC, Andrei 
Alexandrescu wrote:
> Also, there's a more productive view - instead of a 
> classification of ranges, think of "length" as an optional 
> attribute of a range.

On a more abstract level this is indeed a nice view. In case of 
iota though, I'm lost a little bit with this, especially the 
version with "if (isFloatingPoint!(CommonType!(B, E, S)))". 
Depending on B, E and S this might have a calculatable length or 
not. Should we omit length here altogether? If not, how to cope 
with the cases, where we cannot calculate it?