...the range built-in in Python excludes the upper bound. Makes sense from a mathematical point of view (combining different ranges is easier) but I prefer the more intuitive Perl way to do it.

Theoretically, I prefer semi-open ranges [begin, end), aka half-open intervals, because:

• The range size is simply "end - begin"
• Empty ranges are expressed as "begin equals end" and so do not require special handling
• Two subsequences are adjacent means that the upper bound of the one equals the lower bound of the other

This theoretical superiority was eloquently expressed in hand-written notes by Edsger W Dijkstra in 1982, who further argued that zero (not one) is the natural first array subscript, as in [0, N). With typical attention to detail, I see that the three page numbers of Dijkstra's note are: 0, 1, and 2! :)

In practice, I prefer Python semi-open ranges to the inclusive (closed) ranges emitted by the Perl and Ruby range operator. I remember finding Python's semi-open ranges nicer when golfing with string slices. After enjoying Python string slices, I miss them when coding in Perl; the closest Perl equivalent, the substr function, seems unwieldy by comparison.

Semi-open ranges also feel comfortable to me because they form a crucial part of C++ STL, in particular iterators, which in turn were influenced by C pointers and arrays. Stepanov extended some common (semi-open) C idioms, such as:

for (i = 0; i < N; ++i) { // a[i] ... }
for (ptr = a; ptr < a+N; ++ptr)
[download]

inventing a more general iterator abstraction:

for (iter = begin; iter != end; ++iter)
[download]

thus enabling STL algorithms to work on any container that implements the iterator interface.

References

• prof. dr. Edsger W Dijkstra, hand-written notes, "why numbering should start at zero", 11 August 1982
• gcc C++ Iterators, One Past the End
• Stack Overflow question on C++ iterator ranges