On Fri, 06 Feb 2009 11:34:28 +0100, David Clarke <w3@dragonthoughts.co.uk> wrote: > If on the other hand we propose standards where the lack of > normalisation of is tolerated, but require late normalisation, we can > produce a functional result. As they stand, the normalization > algorithms, and checks, are fast to execute if the input is already > normalised to their form. With this in mind, the majority of the > performance hit would only come when non-normalised data is presented. 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.) (Ignoring for the moment the enormous cost and pain of changing codepoint-equality-checks to canonical-equality-checks in widely deployed software and standards...) -- Anne van Kesteren <http://annevankesteren.nl/> <http://www.opera.com/>Received on Friday, 6 February 2009 10:51:18 GMT
This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 6 February 2009 10:51:18 GMT