Yea, I guess in this case back-ward compatibility beats purely correct behaviour. The very best option would probably be a -compatibility switch, but that might complicate things more? If this is the only case where we run into a scenario like this, that might be overkill. But perhaps if we stumble upon more cases we could consider such a switch?
Roland > In my opinion, this behavior of the old compiler is inconsistent and > should be considered a bug. So we have to decide how bug-compatible we want > Falcon to be with the old compiler. It may be difficult to even achieve > bug-compatibility in cases like this. > > Could we live with Falcon compiling it as a string with one space, > regardless of whether there is CDATA or not? >
