Comment 9 for bug 476293

I never use -l N; our histories are so complex that I wouldn't know what revisions to expect from -l N. On the other hand, -rX..Y at least tells me that I'll likely see revisions on the path from X to Y (including merged revisions or not), which is useful (to understand what was added between X and Y).
So I form the wish that -rX..Y is not made slower than it is. I see no problem if it starts printing revisions and ends up with an error message, instead of pre-validating parameters, if that allows to avoid a performance problem (here I refer to John's post of 2010-01-19).