Re: [xml-dev] When writing standards...
If you can replace two formats with one, without breakage or loss of interop, then I say go for it.
#Removing complexity from the world is always good.
#Think of this like code factoring, but on a larger scale.
#
Who is this semi-divine "you" who gets to impose uniformity?
"Removing complexity" is just guff: complexity is not an absolute thing that is unrelated to technical debt, use cases, and so on: the alphabet would be less complex without the letter C, but removing it would be enormously impactful for no benefit. Elegance is good for new things, but standards are frequently not new things, they are messy snapshots of technology in evolution.
There is another law that is co-equal and in tension with Postel's Law: "A Spoonful of Sugar helps the Medicine go down". This Law crops up in Larry Wall's maxim roughly that simple things should be simple to do. But what is "simple" is usually what is "frequent", and what is frequent depends on the use-case, which brings us back to the point that spurning technologies by fiat becomes an exercise in denying the use-cases of others. A standard is an agreement on some technical description developed formally or informally in the light of some explicit and reasonable quality methology, hopefully with some indication of its motivating use case and sweet spot.
I write for people who have brains, like to think, are educated, care about interop. I understand that people reading specs are not computers.
#
Plain English is not good enough, because important permissive/volitional words like "must", "should", "can", "may", "might", "will" and their negatives are ambiguous and even region-dependent. Those, at least, need to be explicit and controlled. Especially where there is legal intent.
Furthermore, more non-native speakers read English than native speakers: the pronoun "it", in particular, is a tarball of confusion. But "plain English" without "it" is just not plain.
So I appreciate the intent (cough XSD splutter), but I think the key is that where simple, formal/executable/lintable ways to express things exist, these should be used: BNF, regular expressions, Unicode character classes, XPath, DTDs, UML, etc and so on, with the helpful English being non-normative peritext.
Regards
Rick
