Language for the ages

On 10/20/2005 at 4:03 PM Sridhar Ayengar wrote: Actually, this dovetails very neatly with the topic and perhaps illustrates the difference in thought between the C/Java/Algol and the FORTRAN camps. It all depends on where you live (mentally). At least up until F77, recursion was not a part of the language--and in fact, was prohibited by many implementations. To a great extent, this hinged on the lack of a hardware-facilitated stack. For many machines, the subroutine call sequence looked like this: EXIT: JUMP *-* ENTRY: first instruction .... JUMP EXIT The machine's CALL instruction would simply store the return address in the location just before the subroutine's entry point. Yes, I know--it's code modification, but it wasn't such a pariah until relatively recent times. Storage local to the subroutine (as well as any memory needed to save the caller's context) was kept in static space. You work with what you've got (try implementing a recursive call mechanism with local storage on an IBM 1620 Model I). Tens, if not hundreds of millions of lines of production code was written this way. Pure recursion was not unknown, but was generally held to be something for the textbooks. While recursion was not generally used, occasionally routines needed to exhibit parallel reentrancy, particularly in the case of event processing, such as I/O and so did need to maintain some sort of stack structure. This sort of thing was usually dreaded by programmers, not because of the inability to code such stuff, but rather because of the difficulty in debugging reentrant code. I believe that this still holds today--and IMOHO the same applies for debugging recursive code failures. On the other hand, an iterative code overflow generally allows for one to make a graceful exit and preserve of the history of the loop. There's rarely any question about "how do I get back to the point of origin?" (i.e. unwind the stack). I remember in a SNOBOL programming course taught by none other than Bob Dewar, the "Tower of Hanoi" example was trotted out. Immediately, one of the students pointed out that it was far easier to write an iterative solution in just about any other language than SNOBOL--and perhaps the real problem was that SNOBOL lacked good iteration facilities. Suffice it to say that Dr. Dewar was not amused. I think this illustrates perfectly how thinking has gradually shifted over the years. Who knows what common practice today will become the b?te noir of tomorrow? Perhaps state machine models will be the next trend and both iteration and recursion will be viewed as hopelessly quaint... Cheers, Chuck