Let me emphasize that at this point, I can be helped most by just getting this file to work!
====================
That said, I'll be happy to keep or change key assignments based on experiences of users (myself at this point) and analysis.
Thanks for the review, Dominik! To address your concerns:
* This is targeted at scripts *Indian* users tend to use: devanAgarI and kannaDa in my case. I use IAST rarely, so I am happy to use whatever is available there (written by you, I believe).
* The main improvement I want over the existing ITRANS-based input scheme is to type the key 'a' less often (about 13/20 times less in Sanskrit). I want to be able to write कमल with just kml, rather than kamala; and शङ्कर with shnakara.
* So there is this focus on efficiency while retaining other ideals such as phoentic intuitiveness and backward compatibility with other schemes.
* I want to retain (less optimal) redundancies to allow compatibility with older, far more popular input methods such as ITRANS, BARAHA, HK; so that interested people can transition without being put off by a (relatively steeper) learning curve.
* ङ् and ञ् are almost always followed by characters of the same varga, such as शङ्कर and सञ्चित. So, one almost never ends up typing ~N and ~n. For example: one just types shnakara and snacita (or snachita) to get शङ्कर and सञ्चित. So, the inefficiency in ~N and ~n is outweighed by its familiarity to ITRANS users. I'm open to adding more efficient alternatives too.
=====================