Where/When did I every suggest that the system is the core design of the next Facebook or Google? I never suggested such a scope, I suggested that when understood the system is easy to extend, modify and maintain, and saves programmer time. I also suggested that given the state of current hardware the extra processing overhead is largely irrelevant, even more so on the next version which I'm working on at the moment.

If I was writing a core for a google or a facebook, then the whole approach would have to be different, I know that... it's not news to me... and wasn't 4 years ago. To get down to 0.003 for your average pagehit you can't be using anything that requires much processing time on the stoneage classical hardware currently available. My point is that 99.9% of people who just want to use Perl to put together a site for their business or community have no need to worry about whether their site is going to become a global hotspot. Such people are looking for something simple they can setup easily and don't have to spend years reading and studying up how to make it fast.

This to my mind is one of the big reasons Perl is losing out to PHP, because PHP is very quick and easy to get results with, and Perl takes effort. People don't like effort, they are either too lazy or too busy to dedicate the time and resources it will take. Especially when they think they can just get an Indian PHP programmer to build what they want for peanuts. And there is no shortage of such people.

What you expect from aXML, a system you know very little of, is a reflection of your own vain ego, it's got to be the biggest/best/fastest to satisfy your craving for power. Not all cars have to be Ferrari's there is room for ford people carriers as well.

This to my mind is one of the big reasons Perl is losing out to PHP, because PHP is very quick and easy to get results with, and Perl takes effort.

In my mind, that's more about deployment than it is about learning syntax. People don't learn syntax. They copy and paste and tweak example code until it appears to behave appropriately.

Plack helps with Perl's deployment. Plack helps a lot.

I suggested that when understood the system is easy to extend, modify and maintain, and saves programmer time.

That'd be easier to judge if you were to provide more examples.

I also suggested that given the state of current hardware the extra processing overhead is largely irrelevant...

The hardware specs you quoted are pretty heavyweight though, especially for a site measuring traffic in the hundreds of thousands of hits per month. When a persistent pure-Perl Plack-based server like Twiggy or Starman can serve thousands of hits per second, the bottleneck gets back to inefficiencies in processing, IO latency, and database traffic.

Even so, moving to a persistent process model is probably the best thing you can do. (I prefer Plack to mod_perl myself these days for many reasons, and not only performance.)