- From: Anne van Kesteren <annevk@opera.com>
- Date: Fri, 06 Feb 2009 15:39:06 +0100
- To: "Aryeh Gregor" <Simetrical+w3c@gmail.com>
- Cc: "David Clarke" <w3@dragonthoughts.co.uk>, "Henri Sivonen" <hsivonen@iki.fi>, public-i18n-core@w3.org, "W3C Style List" <www-style@w3.org>
On Fri, 06 Feb 2009 15:27:26 +0100, Aryeh Gregor
<Simetrical+w3c@gmail.com> wrote:
> On Fri, Feb 6, 2009 at 5:50 AM, Anne van Kesteren <annevk@opera.com>
> wrote:
>> Several people seem to have the assumption that there would only be
>> performance impact for non-normalized data. That is not true, for one
>> because an additional check has to be made to see whether data is
>> normalized
>> in the first place. (And as I said right at the start of this thread,
>> milliseconds do matter.)
>
> Do you think there would be a significant impact on performance if the
> input was just normalized as it's read, instead of being normalized on
> comparison?
I don't know. Developers from Gecko seem to think that is at least
something worth considering, but it does not appear to be what the i18n
guys (for lack of a better term, I'm sure we all care about i18n) want. (I
note I did ask for research here, but nobody so far has come forward with
numbers.)
> Also, it was said that Gecko interns strings -- could it
> normalize them right before interning, so that subsequent comparisons
> are still just pointer comparisons?
That would not help with strings that are not atomized. E.g.
function search(s) {
...
return node.textContent == s
}
or (a feature that might be added to Selectors Level 4):
:contains("...") { ... }
(I might be wrong here, since I do not work on Gecko.)
--
Anne van Kesteren
<http://annevankesteren.nl/>
<http://www.opera.com/>
Received on Friday, 6 February 2009 14:40:03 UTC