That is exactly the problem!

On the N900, it Just Works™. It may not work optimally and may look ugly without some Hildonization, but it works, so you know right away that any effort you are going to make porting it is not going to be completely wasted.

On the Jolla, it Does Not Wok At All™. Not, "does not work well". Not, "looks ugly". No. Does. Not. Work. At All. So you may spend days or weeks trying to port it and all this time you have no idea if you are just wasting your time.

This is not just for porting old code. I am currently writing a new app for the N900 and thought I could build it for Jolla as well. No dice. It builds without problems and even runs, but is completely unusable. The screen rendering is completely off. I have very little time to spend on it, in the order of 1-2 hours per month. I could deal with some #ifs for Maemo and Sailfish but maintaining two completely different code bases with completely different paradigms - forget it.

Besides, I beg to differ on the whole QML business. QML is the spawn of the devil. Just like Glade was. Yes, it makes Sailfish patches easy but as a developer, letting my users fiddle with my application like that is the last thing I want to do. Out of the window goes any idea of security. Jolla must be slapping themselves for making it so easy.

Lastly, QML is a resource hog. No wonder that a mobile phone can hardly breathe with a whole 1GB RAM. One whole gigabyte! I have heard people ask, "how did the N900 manage with that little?" I ask, "what does Jolla do with that much?" QML? Thanks, but no thanks.