Generally it's good idea. dr_frost_dk's charging port connected to battery isn't bad thing, yet, it have some limitations. First of all, it can be charged only by Li-Ion charger of any kind (not power supply). Second, and more important thing, that it should *never* be left charging to 100% during device runtime - power used by device fools Li-Ion charger, which keeps charging battery more than needed = unhealthy for battery life. Last and least important thing, is that bypassing N900's charging circuit confuses bq27x00 chip, so no way for reading 100% accurate (after proper calibration) battery SoC readings (voltage-based reads can be great, but will be never even close-to being such accurate as chip measures).

On the other hand, using port other than semi-fragile USB got obvious benefits. Considering this, I'm more and more tempted to create some crazy solution for full body replacement, that would mean having external charging port connected to testpads. This way, we would have best of both worlds - charging by any 5V power source, without touching USB port, yet still processed by N900 charging circuit.

I can't promise anything, but I'll check what is possible - in meantime, could someone - fluent in reading N900's PCB schematics - check, if for 100%, testpads +5V and mass comes through same safety elements as USB charging? I'm sure about USB data lacking only common-mode choke, but about power lines, it's unverified info for now.

/Estel