From: John B. <joh...@ne...> - 2000-08-25 13:21:57
|
On Thu, 24 Aug 2000, Graham Barr wrote: > However I am thinking, is this really worth it ? According to the > above the parse is taking 0.122 seconds and the load takes 0.028 a > saving of 0.1 seconds at startup time. Even you times only said it > saved 0.24, is this extra maintenance (ie any change to the parse tree > will need to be accounted for) really woth that extra 0.24 seconds ? I'm not sure. On my home box I found it took around 60% of the startup time (for Net::LDAP::ASN) that it did previously. It is only of use (if at all) for those one-shot applications where command latency is an issue. This is something I've noticed in the past being a little annoying. For me, this 'feels' snappier when I run one-shot commands which use Net::LDAP. The potential maintenance burden is not something to be taken lightly, I agree. It might be possible to code a little more defensively against ASN.1 tree format changes and/or have a fallback method to parse the definition if the load fails. I don't know if that would help or not. What do other people think? I believe that there was some discussion a few weeks back about Net::LDAP seeming much slower than C APIs when used as a command line tool which I believe boiled down to startup times. Would anyone involved in that care to try this patch and see if there is a significant improvement for their application? If the current format of the patch is problematic to employ, mail me directly and I'll provide a more easily tried version. regards, jb |