I wouldn't say it's impossible. You could have as the rule something
like code xyz becomes fu'ivla xaybzc, where a, b, and c are functions
of x, y, z such that xaybzc always results in a valid fu'ivla. The
problem is that the result would usually be hardly anything
recognizable.
> I was trying to keep the CCV'VCV structure for all of them....
Even more restrictive.
True, but also more internally consistent.
>> > Assamese (অসমীয়া) sma'ine (asm/ine)
>> > Belarusian (Беларуская) ble'ine (bel/ine)
>> > Breton (brezhoneg) bre'ine (bre/ine)
>> > Chechen (нохчийн мотт) tce'acu (che/cau)
>>
>> OK, don't let me stop you. :)
>
> Well, it's fun and all,
Was your list picked randomly, or were the examples hand-picked so
that they would convert easily into a fu'ivla? If the former, you may
have something, if the latter, the method has not really been tested.
Something in between. I had only done those three language families, or maybe one or two more. So they're not entirely random, but they're also not really hand picked.