[This is a mail reply I sent to someone who wondered what I meant by ``this behavior is mentioned as a `common extension' by the ANSI Standard, no pun intended.'']

I guess it's sorta obscure if you don't know too much about FORTRAN.

In FORTRAN, there were data structures called ``COMMON blocks'' which were, as I recall, the main way of sharing global data between the modules of a separately-compiled program. They were a nuisance to use: you had to mention the same variable names, in the same order, in the COMMON directive at the top of each source file. (Actually, you could use deliberate mismatches of variable names or types in COMMON blocks to inspect the bits of a value of one type as if it were another, much as you might do with a union in C.) When you added another variable to a COMMON block, you had to remember to add it to all the definitions of that COMMON block, in all source files. If you'd never heard of preprocessors or source file inclusion (that is, of mechanisms like C's #include) you invented the idea for yourself, and wished that it existed in FORTRAN. (It didn't, at least not in the standard versions of FORTRAN when I used it, although some compilers -- probably including Unix's -- supported it as an extension.)

That's by way of background. In C, there's always been some uncertainty about the precise way global variables are handled, which is of course why FAQ 1.7 exists. In particular, in the classic implementations of C, it was possible to define a global variable like 

	int i;
in several source files, and as long as no more than one of the definitions had an explicit initializer, the linker would realize that they were all the same variable, and essentially overlay them in memory. This was referred to as the ``common model'', not because it was popular or widespread (though it was), but because it was reminiscent of FORTRAN COMMON blocks.