> I can see the nil being returned there explicitly, so that's "where it's happening." Should this be the case though?

The previous, probably original implmenetation of #peekBack used to send
#oldBack to the stream.
IIRC #oldBack was based on the idea that the position of a stream is an
index of a sequence where #next means +1 to that index while #back means
-1 to that index. Following that logic, you have to #skip: -2 and send
#next to get the element -1 to position.

#oldBack has been removed but the behavior of #peekBack is presumably the
same as it was before. Some ancient but now external code may rely on
#peekBack but it's not very likely such code would work in the current
#peekBack has no real users in the Trunk only a test remembers what it
used to do.

Given the above, and given that we have #next:, I ended up balancing that API with #peek:, and that again with #peekBack:.  Since they all simply return a String, possibly empty, they dodge the question about nil vs. error entirely. 

So, I think it's a good time to change its behavior to be based on #back.

Without wanting to sound ungrateful for Eric's contribution (of which I look forward to more of), may we also consider the addition-by-subtraction opportunity?  I mean, it kinda makes sense that there wouldn't, and won't, be any users of #peekBack.  Maybe we should deprecate it.

 - Chris