(preliminary, very much a work-in-progress.)

Tired of being called a dummy?

Note: This is not a beginner's tutorial.  If you do not already have a good grasp of programming in general and C syntax, this is probably not the place to start.

A few words about history, ANSI and ISO

The C language grew out of a relatively long-term effort at AT&T Bell Laboratories.  The first book about C, The C Programming Language by Brian W. Kernighan and Dennis M. Ritchie, was first published in 1978.  This book is often called ‘K&R’, after the authors' initials, or the ‘White Book’, a nod to the spare white cover.  The language was by then already rather popular in certain niches, and with publication of the White Book, spread rapidly.  In the early 1980s, the American National Standards Institute formed a committee (X3.159) to standardize the language.  The committee added a few features and made a few ‘quiet changes’ that broke existing programs, but on the whole stuck relatively closely with the original language.

The first ANSI C standard was published in December 1989, and was thus numbered X3.159-1989.  The International Organization for Standards (ISO) immediately adopted the ANSI C standard, making only some editorial changes and giving it a new number, ISO 9899:1990.  This version of C was thus variously called either ‘ANSI C’ or ‘ISO C’.  The older variant was often called ‘K&R C’ or ‘White Book C’, or -- after the publication of a second, updated edition -- ‘K&R-1 C’, referring to the first edition of the White Book.

During the 1990s, however, ANSI and ISO proceeded to update these standards, producing both a set of ‘Normative Addenda’ and then eventually a new standard, ISO 9899:1999.  By prearranged rules, this was automatically also X3.159-1999, supplanting X3.159-1989.  This means that there are actually two versions of ANSI/ISO C.  Here I use the name ‘C89’ to refer to the old standard (many prefer ‘C90’), and ‘C99’ to refer to the new one (this time there is only one year-number).

Most of this document focuses on C89.

Standards?  We don't need any stinkin' standards!

Well, actually, maybe you do.  It really is a question of what you want to do.

The standard -- either one -- is not the End of All C.  Writing ‘strictly conforming’ C code, however, has an enormous benefit.  The C standards are, in essence, a contract between you -- a programmer writing C code -- and the implementor supplying your C compiler and library.  This contract says, in effect, that as long as you stick to your part of the bargain, the implementor must stick to his.  The moment you violate your side of the deal, the implementor is free to violate his, too.

In practice, this means that any strictly conforming code you write will work on every C compiler.  (A compiler cannot really claim to be a ‘C compiler’ unless it implements Standard C.  There may be no Standards Cops to haul the implementor off to jail, but in fact, if you wrote code that failed to run properly, and someone was hurt as a result and that person sued you in court, your legal liability might depend to some extent on whether you followed ‘accepted industry practices.’  If the court were to decide that those practices included coding in Standard C where possible, one of the key legal points in deciding whether you were ‘negligent’ -- in the legal sense of the word -- could hinge on whether you coded in Standard C.  Of course, simply adhering to Standard C might not be sufficient, and in a court case, there are many variables, but being able to say ‘I used accepted standards and practices in my work’ has direct legal benefits in certain types of lawsuits.  Please note that this is not intended as legal advice; I am not a lawyer; I just listen to them.)  More often, it also means that you can change compilers or even machines.  If, after getting halfway through a program, your boss comes to you and says ‘the model Q computer is obsolete; we're buying model R machines now’ and you wrote Q-specific code, you will have to rewrite it all.  If you wrote standard C, it will all work on model R.

Standard C is quite limited.  It has no networking, no graphics, no mouse support; it cannot start processes, print to printers, and so on.  Nonetheless, it can do a lot -- and if you decompose whatever program you are writing into ‘portable’ and ‘non-portable’ sections, you may never have to rewrite the portable parts.  If you need to start a process over the network, put that task in a separate function in a separate file.  Supply a different function (and file) for each different system as needed.  Your program will still require network and process extensions, but now you will only have to rewrite the ‘start network process’ function each time you move the program.



Gilbert & Sullivan would roll over in their graves