Why do we allow 0x descriptors?

[yoavweiss]

It seems that the x descriptors are defined as non-negative integers. Shouldn't they be positive integers?

[yoavweiss]

I got it very wrong - x descriptors are floats!!!
Still, can we define them as positive floats?

[eeeps]

I'd be in favor of requiring x descriptors to be positive floats, and w descriptors to be positive integers. I don't see how either could be reasonably set to zero.

[zcorpan]

Per irc discussions, 0x is basically like <img width=9999999999999999999999999999999999> which is allowed. We should allow it and require an infinitely big image to be rendered, and implementations would have a limit (allowed by http://whatwg.org/C#hardwareLimitations ) on how big images can be. Normatively we don't need to add anything, but we could add a note somewhere saying something -- any suggestions?

For 0w, we could drop it during parsing, but I don't see any benefit in doing so and there's a risk of introducing bugs with changing the algorithm. However we can basically disallow it by requiring that it match the image's actual width.