From: Kurt D. Z. <Ku...@Op...> - 2000-08-25 22:19:00
|
At 02:21 PM 8/25/00 +0100, John Berthels wrote: >What do other people think? Well, since you asked... I think it's a little premature to be optimizing net::LDAP. I'd prefer a complete and correct API over an optimized one any day. And, when "completeness" is declared, the optimizations that will be critical to me will improve overall runtime performance (such as Convert::ASN1 optimizations). As you've noted, start up time only matters for small "one-shot" applications. I would argue that such applications which were also time critical would be reimplemented using a language/API that compiled native executable. For my uses of net::ldap, start up time rarely factors in. Even for small "one-shot" applications, they are usually just a little script I just whipped up to fulfill a requirement of the moment. The startup cost of net::ldap (and perl in general) is easily offset by other costs (such as my time whipping the thing together). And if this script ever became time critical, I'd write the thing in C to avoid the perl overhead or use another approach (such, as in the web space, changing from CGI to mod_perl). Now, I'd like to see the completed net::LDAP optimized. I believe net::LDAP can be optimized to operate at comparable run time speeds of other Perl LDAP modules... and startup time (if it's really an issue) could be improved as well... but completeness and correctness first. Kurt |