dsill@NSWC-OAS.arpa (Dave Sill) (04/14/88)
[I hate to add fuel to the fire...] Nevin Liber writes: >It still IS dependent on the [implementation]; you just want ANSI to put the >implementation in the specification. > >You are not defining *what* the function does (ie, you are not making an >abstract *description* of the function); you are defining *how* the >function does a strcpy (ie, how it is [supposed] to be *implemented*). This is simply not true. Defining strcpy so that it would work right for overlapping strings WOULD NOT DEFINE HOW STRCPY IS IMPLEMENTED. There would still be an unlimited number of ways to code strcpy. I support such a definition because it would break NO existing code and it would codify existing practice. So who loses? People on machines whose most efficient block copy is address-descending, maybe. Perhaps 5% (I'm being generous) of the C-using population. And who wins? Everybody else who uses or maintains code that made this non-portable assumption. ========= The opinions expressed above are mine. "It's said that premature optimization is the root of all evil." -- Bill Joy