Dave Winer wrote:

If you’re going to do an API please use XML.

…and I am doing it for a project upon which I am working, but I am increasingly asking “why”?

There’s nothing magical about having extra angle brackets, irrespective of the beliefs of the Church of SGML.

I think the answer is in part given by a commenter in Dave’s thread: (my emphasis)

If you are doing an API that will be accessed from a browser use JSON. XML is the wrong answer. But most people ought to just offer both. It is trivial to offer your API in several formats. – Sam Pullara

– it didn’t used to be that way. Time was that serialisation / transport encoding was hard – if not something of a a black art – APIs were closed, and programmers were coersced to use overengineered “standard” encoding schemes that would have to cope with 9-bit bytes on occasion.

That battle is over. Bytes are 8 bit, words are multiples of bytes, and openness prevails, mostly. It’s even OK to toss integers around as strings, since surely there is so much compute power flying around that the waste of bandwidth (bitwidth?) is not a problem?

So now we’ve gone the other way and engineers are coersced to use overengineered <SPAN TYPE=”bulky”>bulky</SPAN> encoding formats in the name of the forward progress of the people, huzzah!

Except that’s crap, too. Why should I have to implement a sodding enormous parser within a scripting language just to toss a few integers around? Don’t say “Atom” and stuff, because that’s the tail wagging the dog.

So I would fix Winer’s request:

If you’re going to do an API please use an open standard; and open-source your code so we can fix any brain damage that you might miss. Use XML where you’re dealing in some sort of document interchange, otherwise use what’s popular.