ext Sagar R. Shah ([EMAIL PROTECTED]) wrote: > > Why does it 'make sense' to deviate from pure perl? I'm not disputing that > doing it in C would probably be faster, but you seem to be implying that > there are not enough pure perl optimisations that can be done to get a > comprarable speed-up.
Let's take the case of my X.509 certificate parsing module. Parsing one certificate with Convert::ASN1 is ~25 times slower than doing it with OpenSSL. I like it, though, since I can run it on any platform which perl runs (which may not have OpenSSL). > Are you pretty certain that pure-perl optimisations/changes cannot get a > comparible speedup? I would bet against it, but I wonder what Graham has to say. > Ditto :) > Out of interest, do you use another module in the other 10% of cases? Or > just take the performance hit? I don't use Mozilla::LDAP at all, only Net::LDAP. In the other 10% of the cases, I write the code in C, using either OpenLDAP or Netscape C LDAP SDK. These cases include any command-line programs which will be executed thousands of times per day, because each invocation loads a big, fat perl interpeter into memory and drives the system load up. The exception is if you happen to use mod_perl... BR, -- Mike Jackson
