I think the rules changed at some point by adding this castling business, and that the applets play by the old rules, where a castle always decomposes. Apparently it was perceived as a flaw that castles could not be salvaged from check, or forced to decompose by zugzwang, and castling and stalemate were introduced as fixes.
I think this was a poor choice; the castling destroys the conceptual simplicity and elegance of this variant. If integral motion of a royal piece was desired, the logical solution would have been to make the king one of the atomic pieces. E.g. if de rings are designated L(arge), M and S, we could have L=K, M=B, S=R, LM=Q, MS=N, LMS=P. If you start with 8 LMS pieces, the victory condition could be that having no kings at the end of your turn loses. Your first move will create a K (leaving an N behind). A K then cannot decompose, and can capture a Q or P. It could turnover a B or N, but perhaps the rule should be that K cannot turnover, and only capture orthogonally.
I think the rules changed at some point by adding this castling business, and that the applets play by the old rules, where a castle always decomposes. Apparently it was perceived as a flaw that castles could not be salvaged from check, or forced to decompose by zugzwang, and castling and stalemate were introduced as fixes.
I think this was a poor choice; the castling destroys the conceptual simplicity and elegance of this variant. If integral motion of a royal piece was desired, the logical solution would have been to make the king one of the atomic pieces. E.g. if de rings are designated L(arge), M and S, we could have L=K, M=B, S=R, LM=Q, MS=N, LMS=P. If you start with 8 LMS pieces, the victory condition could be that having no kings at the end of your turn loses. Your first move will create a K (leaving an N behind). A K then cannot decompose, and can capture a Q or P. It could turnover a B or N, but perhaps the rule should be that K cannot turnover, and only capture orthogonally.