The framework I began several months ago is certainly showing its shortcomings, and I am in the process of completely gutting it.

The general premises that have lead me to the conclusion this is needed are listed below:

1) All business logic and database logic is contained within one file in one class, from my own personal experience I believe it would be far easier to maintain if I move database logic and business logic into seperate classes, supporting the MVC pattern.

2) The actual validation methods are in procedural functions. There are two ways that I am looking at rectifying this. Either I move all validation into a seperate stand alone validation class and load that class when it is necessary to validate, or I build all validation into the base controller class, which means more overhead.

I am currently leaning towards option 1.

3) The views are currently split out into a set of templates which get parsed and put together. The main issue with this one is speed. It takes time to run through each script tag with a regex (and sometimes multiple times if the script includes another script). So the options are to scrap the scripting language entirely and revamp it to use the base language (php) which would fix the speed problem but would muddy the views a bit and could potentially cause a bit more work for me. Alternatively I could just revamp the current engine, the only real downside is that I may not be able to come up with a sutible alternative and end up with option 1 and a ton of wasted time.

More to come as it gets closer to release.