On 8/12/13 2:37 PM, "Carsten Bormann" <[email protected]> wrote:

>        If section 3.6 stays, the numerics need more work.  +/- Infinity
>should be
>        treated like NaN.
>
>(Why?)

Having more than one way to encode +/- Infinity seems like a recipe for
generating slightly different canonical output.  For example, in my code
right now, I think I'm always generating a half-precision Infinity, and
I'm thinking about changing that to always be the 4-byte version.

I now see that this section is just a recipe for how to define a
canonicalization, not the actual rules you need to follow, and my comment
may be moot.  However, I suggest that section 3.6 is therefore not
terribly useful, and it might be removed without harming the document.

>        In Appendix D, shouldn't the input be unsigned or an array of
>bytes?
>
>It could also be unsigned short (which would be widened to int
>anyway), but in this case it shouldn't matter.
>(Array of bytes would just be slightly more tedious.)

Agree that it's merely a tedious transformation, but I think it would be a
lot more clear.  Semantically, the input isn't an integer of any size.
It's two bytes in a specific order.

>        Add PER.
>
>PER does need schema information even for parsing and would require
>its own little dissertation, so we left this off.

I think it's useful to call out an example that is explicitly
schema-ridden, particularly since it has been called out in this
conversation as something relevant.

-- 
Joe Hildebrand



_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art

Reply via email to