I think our user base is insignificant to what it will eventually become and we need to address this as soon as possible to avoid endless confusion. Better now than later.
I think that the demands on the string class are such that a general purpose string will not be acceptible. From the earlier design discussion, it was mentioned that the code attempts to identify repetative strings and only have one copy in memory. That seems to be an optimization that we don't want to lose. The current behavior is mysterious to me, maybe I'll have a chance to explore it and familiarize myself with Java's string buffer and come with with some minor tweaks that prevent the node changing affect described previously or at least document the behavior.
