The first time I used Hal 5 I was very interested, but I noticed 3 main weak points, which are now completely solved by Hal 6:
1. If you teach Hal 5 many things, you get hundreds or thousands of scattered files, which is not a good strategy for a wide information system. Hal 6 uses tables included in one file.
2. Every time you want to add some functionalities to Hal 5, you must edit the main brain; again, a good information system must keep the original coding separated from the customized versions, so that it is easier to activate/deactivate any customer solution, and to apply subsequent upgrades/patches to the main code, without affecting the customized code. The plugin architecture is really a very good approach to this issue.
3. In Hal5 there was no way to access the brain files according to some standard language. Hal 6 uses SQL, which allows to write software based on the most used standard query language.
All these aspects are not evident from a user point of view, in the normal interaction with Hal, but they are strategic platform changes, from a software architecture point of view, which open up Hal to significantly different expansion options.
A new version may contain some more bugs than an old consolidated version, but I think that Rob's ability, together with some little help from all us in debugging and testing, will overcome them.
So I am strongly committed to Hal6. My only disappointment is that I had written some plugins for the Haptek full body character, linking the display of emotions to the conversation flow, and now I have to use that horrible frog instead (sorry, Rob Med, I believe you are exceptionally talented as a software developer, but your graphic tastes are at least questionable). I hope that the Haptek plugin will be available very soon.