As you’ve read Core Data vs SQLite 3, you know that Core Data and the persistence mechanism (SQLite in this case) are largely orthogonal. Core Data is really about managing an object graph and it’s main use case is for the model component of an MVC architecture. If your application fits nicely into this architecture, it’s probably worth using Core Data as it will save you a lot of code in the model component. If you already have a working model component (e.g. from the existing desktop app), then Core Data won’t buy you much. A hybrid approach is possible– you can do your own persistence/querying and build a Core Data in memory store which you populate with the result of a query and use this in-memory store via Core Data as the model component for your app. This isn’t common, but I’ve done it and there are no major roadblocks.
To answer your specific questions:
You can assign a version number to the entire persistent store and retrieve that information via
+[NSPersistentStore metadataForPersistentStoreWithURL:error:], without even opening the store. An equivalent
+setMetadata:forPersistentStoreWithURL:erroralso exists, of course. If you want to store the version info in an entity instance instead of in the persistent store metadata, you can load only a single object. With an SQLite persistent store, Core Data does a very good job of fetching only what you need.
NSPredicateAPI, is very easy to learn and it seems to do a decent job of compilation to SQL. At least for databases of the size you could fit on an iPhone it’s certainly been adequate (performance wise) in my experience. I think the SQL vs. Core Data question is slightly misguided, however. Once you get the result of a query what are you going to do with it? If you roll your own, you’ll have to instantiate objects, handle faulting/uniqueing (if you don’t want to load the entire result of a query into memory immediately) and all of the other object graph management facilities already provided by Core Data.