[whatwg] Endianness of typed arrays

On 3/28/12 12:59 AM, Charles Pritchard wrote:
> I've no desire to coddle low-level coders. They know what they're
> getting into.

You're making the mistake of thinking that people using typed arrays are 
"low-level coders", whereas in reality everyone and their mother is 
being told to use the arraybuffer response type on XHR to load all sorts 
of data.

And it's not a matter of "coddling".  It's a matter of writing specs 
that are actually useful.  Right now, a big-endian UA that implemented 
the typed array spec as written would be pretty broken on a number of 
sites, because people are using typed arrays for all sorts of things 
outside their original WebGL context.  As typed array support becomes 
even more common in browsers, this problem will only get worse.

The general position of this group, indeed the main reason for its 
existence, is that specs should reflect reality and that a spec that is 
deliberately misleading in terms of what it takes to achieve 
interoperable behavior is not much better than no spec at all (and is 
arguably worse than no spec at all).

You are, of course, entitled to disagree, but I would like to understand 
the reasons why you think deliberately misleading specs are a good idea...

-Boris

Received on Wednesday, 28 March 2012 01:05:16 UTC