- From: David Hyatt <hyatt@apple.com>
- Date: Tue, 26 Oct 2010 13:15:23 -0500
- To: fantasai <fantasai.lists@inkedblade.net>
- Cc: koba Mobile2 <koba@antenna.co.jp>, John Daggett <jdaggett@mozilla.com>, www-style@w3.org, howcome@opera.com
On Oct 25, 2010, at 8:12 PM, fantasai wrote: > On 10/25/2010 05:52 PM, koba Mobile2 wrote: >> John Daggett wrote: >>> koba Mobile2 wrote: >>>> Håkon Wium Lie wrote: >>>>> Implementation cost is one metric, but no the only one. For one, >>>>> there's also a memory cost. I presume you have implemented this so that >>>>> the logical propertie cascade and inherit separately? on a per-element >>>>> basis? And cannot be resolved until you know the computed value for >>>>> 'writing-mode'? If so, the memory use will be significant: ~35 >>>>> property values for every element. >>>> >>>> Your opinion is an excuse from an incompetent implementor. >>> >>> No, it's the concern of an implementor who has a long history >>> of developing browsers for devices that run with restricted memory. >> >> All browsers are not asked to support all properties. > > No, but desktop browsers and their equivalents are. > > That said, the performance and memory considerations for logical > properties are not as bad as they seem at first glance. Since > you can compute the logical-physical equivalence at cascade time, > you only need to store one set of data per element (physical or > logical, depending on your layout architecture). So the extra > memory load is almost nothing. > > I could try to convince you that this is true, but I don't have > to: Hyatt already implemented it this way and can confirm that > this is the case. Yeah the memory/perf cost is virtually 0. The only extra memory consumed is in the back end declaration storage for rules, and then only if you have to specify a physical and logical for backwards compatibility. Aside from the alternate stylesheet solution, all other proposed solutions share the same cost (two physical declarations vs. one physical/one logical declaration). That is pretty inconsequential. In terms of additional memory overhead per-element, there is none. So yes, from an implementation perspective, the logical properties are trivial. dave (hyatt@apple.com)
Received on Tuesday, 26 October 2010 18:15:57 UTC