A quick note about standards

[Andrew Sayers]

A couple of times in recent discussions, It's been asserted that 
standards should be followed at all times.  Without commenting on 
specific cases, I'd like to explain why it's not generally that simple.

A standard is the textual equivalent of a programming library: a tool 
created by a lot of smart guys with significant expertise in their domain.

Like libraries, some standards are universally adopted (zlib, cargo 
containers); some are rejected (libesd, OSI); and some are the subject 
of endless flamewars (widget sets, power plugs).

Also like libraries, anyone can write one and nobody agrees on what goes 
in one.  For example, Ecma believes that a standard should be an 
accurate description of current practice, whereas ISO believes that a 
standard should be a consensus-based reflection of national interests.

Standards should certainly be followed in many cases.  For example, 
you'd be as much a fool to reinvent Unicode as you would be to reinvent 
libgcrypt.  But there's a constant need to evaluate the appropriateness 
of individual standards to the task you're working on.

     - Andrew