Re: [css-counter-styles] 2^15 limits

On Wed, Feb 20, 2013 at 5:21 AM, Peter Moulder <peter.moulder@monash.edu> wrote:
> If we're allowing for implementations to use a 16-bit int to represent a bound,
> then how would a bound of 'infinite' be represented?  I suggest we allow
> implementations the option of using 16-bit int values -2^15 and 2^15-1 for
> this purpose, and thus suggest narrowing the required supported range to
> [-2^15 + 1, 2^15 - 2].

Given that impl limits clamp your range anyway, you just implement
'infinite' as the largest/smallest number you support anyway.  There's
no need to record it as a special value.

> That said, I vaguely recall that 2^15 had some other significance, perhaps
> being the range of counter values that implementations had to support in CSS
> 2.1 or something.

Not really.  We just needed to come up with some bound, and I already
knew that implementations don't quite use a full 32 bits for it, so we
went with the lower bound.

> Apart from that, there are also a few nits in this paragraph; nothing
> important, but most of them look easy enough to fix:
>
>   - Literally speaking, 'infinite' is "a bound greater than the
>     implementation's supported bounds", but we presumably don't want it to be
>     "treated as specifying the implementation's maximum supported bound."
>     I suggest inserting the word ‘integer’.

I don't understand the problem (that is, in fact, exactly what it
means), nor what your solution is.

>   - "If a range ... then it must be treated as specifying the implementation's
>     maximum supported bound" suggests that it's the range that should be so
>     treated, rather than just the bound.  Not a big problem, but it's easy to
>     fix: I suggest removing the "range" part of that sentence, i.e. just "If a
>     bound is specified...".

Done.

>   - It would be nice to find a wording that more obviously conveys magnitude,
>     at least for the "maximum supported bound" part (which fairly strongly
>     suggests just the upper bound, on first reading), and to a lesser extent
>     also the earlier "lower bound of at least" part.  That earlier one is
>     actually the easier to fix, e.g. "at least as negative as".  Maybe the
>     other could be something like "is outside of the supported bounds, then it
>     must be treated as the closest bound that the implementation does support".

Done.

~TJ

Received on Wednesday, 20 February 2013 22:39:04 UTC