It would be possible to become angry about Shmuel's literalisms, selective quotations, context mixing and switching, and the like. It would even be possible to enter into protracted disputes about them with him, as I used to do too often.
I no longer judge it worthwhile to do this. His gadfly posts give him great satisfaction; they are sometimes useful; and even when they are not they are, finally, innocuous. This time, for example, he has, a little opaquely, disentangled the source-language syntactic form of a function reference from its implementation, which may sometimes usefully be in line instead of by library-subroutine call. It is true, for example, that for such PL/I constructs as declare vname character(46) varying, vnl signed binary fixed(15,0), length builtin ; . . . vnl = length(vname) ; IBM and many other PL/I compilers implement the BIF length in line. (There would be scope for C compilers to do similar things in some cases, but I am not aware that any does them.) Still, this distinction between syntactic form and implementation strategy and the option it makes available to compiler writers are important. This case is not an isolated one. It would not be unfair, I think, to characterize Shmuel's posts as a mixture of gratuitous contention and redeeming technical information. --jg ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
