Jonathan Vanasco wrote:
i specifically wanted to keep apache related stuff out of the model when designing it, but i was in the situation where i needed to access it for a quick hack ( i chose the longer elegant solution that negated touching it ) and said "hm... i wonder how could i do this"
It really hard work - I built a framework which used one instance of its main module per Apache instance, and the handler would then call methods on it (so it could implement caching and so on). The following were big issues:
- a file upload module which of course needed access to the Apache2::Upload object - several modules which create URLs needed document_root and other request info - cleanup - each module in my framework needs to refer to the main object and so stores a reference to it, but this makes garbage collection nearly impossible. I've seen a solution to this in the SOAP::Lite modules but haven't had time to find out how it works - I had to get very intimate with the Apache2::SOAP source to talk to a .NET SOAP client.
As an ex assembler programmer I try not to use abstraction for abstraction's sake, so decided that my framework would only have to work with Apache2/MP2 or stand-alone. Made life lots easier.
I also asked myself afterwards why we keep reinventing the wheel! John
