- From: Bjoern Hoehrmann <derhoermi@gmx.net>
- Date: Tue, 07 Sep 2004 04:12:57 +0200
- To: Nick Kew <nick@webthing.com>
- Cc: QA Dev <public-qa-dev@w3.org>
* Nick Kew wrote: >> Of course, we'd need to measure the effects before committing to that course; >> I've always suspected that our obvious gains from mod_perl would be >> insignificant compared to the fork() and processing of «onsgmls». > >I wouldn't be too sure of that, for two reasons. One is onsgmls, which >is not slow for such a complex parser. The other is the generic API: >has anyone benchmarked the Perl bindings thereof? Sure, but you cannot really compare SGML::Parser::OpenSP to onsgmls as the performance is relative to what you use it for. On my system, processing the XHTML 1.0 SE Recommendation and ignoring all output S::P::O should be about 30% faster than `onsgmls ...`. S::P::O's performance drops significantly if you let it create Perl data structures for all event data, if it does that for all events in the beforementioned case `onsgmls ...` would be five times faster than S::P::O minus whatever it takes to post-process onsgmls' output.
Received on Tuesday, 7 September 2004 02:13:52 UTC