I know this has been discussed elsewhere but I am curious about others experience with the separation model. I've used it for some relatively moderate sized projects (thousands of records, 20 - 25 tables, 3 or 4 files). It certainly has appeal for FileMaker Go and WAN access. I've read where some have advocated an absolute minimum be in the data file other than the base tables. Value Lists and everything else should be defined in the UI file. My question is:
With value lists unavailable and no relationship information in the data file, how do you enforce validation other than through robust scripting?
Do you banish most calculations to script generated data that places values in generic Number and Text fields?
With the above, I can see significant performance enhancement (especially with very large tables) but also danger of not having current data unless things are VERY carefully scripted.
What is the recommendation from those of you more seasoned at this? I really want to put things together in an extensible and robust way.