- From: Ernest Cline <ernestcline@mindspring.com>
- Date: Tue, 16 Mar 2004 14:20:32 -0500
- To: "Jungshik Shin" <jshin@mailaps.org>, www-style@w3.org
> [Original Message] > From: Jungshik Shin <jshin@mailaps.org> > > Ernest Cline wrote: > > > cjkv-ideographic <string> <string>? > > This is a generalized version of the "cjk-ideographic" > > 'list-style-style'. The 'v' got added both because CJKV > > .... > > > The first string would be required to contain 16 characters, > > if it did not, then the list style would be treated as if it were > > "decimal". The sixteen characters would be in this order: > > ... > > > Second Group Marker > > Third Group Marker > > Fourth Group Marker > > > > The range of values that could be represented is from > > 1 to 9,999,999,999,999,999. Values outside that range > > would be treated as if the list style were decimal > > CJKV ideographic system can represent numbers up to 10^72-1 :-) See > > http://lists.w3.org/Archives/Public/www-style/2003Apr/0063.html While going past the 4th Group Marker may be correct, it would be hard to implement. There are values using the 5th Group Marker that cannot be represented using a 64 bit signed integer. Some values using the 4th Group Marker cannot be represented by an IEEE double. There are values that use the 3rd Group marker that cannot be represented using a 32 bit signed integer These limits are significant because in the HTML DOM, the value for list markers is of type long which translates to a 32 bit signed integer in Java and an IEEE double in ECMAScript. Hence anything beyond the 4th Group Marker is worthless for current UAs and even with a 64 bit integer, worrying about anything beyond the 5th Group Marker is pointless. CSS 3 Generated Content needs to specify the minimum acceptable range for counter values. Given the values used in HTML DOM, the range -2^31 to 2^31-1 (32 bit integer) seems like a reasonable minimum. (There is nothing I can see in the DOM that would prevent using a greater range for CSS counter values if desired, but requiring a greater range would likely cause some otherwise conformant implementations to be non-conformant for little gain.)
Received on Tuesday, 16 March 2004 14:20:33 UTC