26901 2014-09-25 10:25:24 +0000 Stop using "octet" 2014-12-05 16:45:46 +0000 1 1 1 Unclassified WebAppsWG WebIDL unspecified PC All NEW P2 normal --- 1 annevk cam bzbarsky d jackalmage mike public-script-coord travil w3c public-webapps-bugzilla oldest_to_newest 112170 0 annevk 2014-09-25 10:25:24 +0000 People are justifying using "octet" as terminology because IDL has it. Per https://encoding.spec.whatwg.org#byte we standardize on byte. So... Perhaps avoid calling these bytes directly altogether, as we have ArrayBuffer and friends for that. "short short" and "unsigned short short"? 115946 1 annevk 2014-12-04 18:37:00 +0000 We should just adopt the ES terms here. Int8, Uint8, etc. See http://people.mozilla.org/~jorendorff/es6-draft.html#sec-typedarray-objects 115947 2 jackalmage 2014-12-04 18:38:53 +0000 +1, the ES terms are easy to understand and pretty standard. 115948 3 travil 2014-12-04 18:49:01 +0000 (In reply to Tab Atkins Jr. from comment #2) > +1, the ES terms are easy to understand and pretty standard. +2 :-) 115950 4 d 2014-12-04 18:54:46 +0000 I dunno, I'd rather do something more ambitious (we have another bug for it) where the only type is "number" and you use a restriction specifier for parameters like foo(ClampBetween(0, 255) number bar) // clamps to range baz(EnsureBetween(0, 255) number bar) // throws outside range 115960 5 w3c 2014-12-05 00:44:29 +0000 Elsewhere [1], I proposed something like: typedef Number[−128..127] Byte; Having that would shrink the WebIDL spec significantly too ;) [1] https://github.com/heycam/webidl/issues/14 115961 6 jackalmage 2014-12-05 00:55:31 +0000 "Byte" is ambiguous - it means 8 bits, but doesn't specify how to interpret it. Int8 and Uint8 have no such issue. 115962 7 w3c 2014-12-05 02:55:30 +0000 (In reply to Tab Atkins Jr. from comment #6) > "Byte" is ambiguous - it means 8 bits, but doesn't specify how to interpret > it. Int8 and Uint8 have no such issue. for sure, the point is just to have some syntax to declare ranges... we might be able to kill a whole bunch as they are probably not used in specs. 115966 8 bzbarsky 2014-12-05 08:32:35 +0000 > we might be able to kill a whole bunch as they are probably not used in specs. The following are used in specs as function arguments or writable attributes: octet, short, unsigned short, long, unsigned long, long long, unsigned long long. So the only unused one seems to be "byte". It's used for return values and constant values, but I don't see any obvious coercious to "byte" offhand. 115971 9 d 2014-12-05 16:45:46 +0000 One problem with using the uint8 name is that in theory ES wants to add native int8/uint8/etc. types: https://github.com/nikomatsakis/typed-objects-explainer/blob/master/core.md#primitive-type-definitions. It would be confusing for WebIDL to mean something (very) different. Still it'd be nice to fix the problem in the OP, without having to first solve the whole ambitious reform-number-types idea. Not sure what to do. (I plan to go comment on Marcos's GitHub issue with ideas for number types.)