Hi Tammo, > I'm not sure if your suggested patch is sufficient in all cases. If I > understood Xerces' DOM model correctly, then it appears to be not > thread-safe, not even for read-only access. In this case just forcing a > pre-initialization of the lazy model is probably not enough. > > I'm currently thinking about cloning the element for those parts that > may concurrently access the node. How does that sound?
As long as it's not an performance killer, any solution is fine for me (-; > Is it possible to easily reproduce the problem in a test case? Or do > you have some kind of testbed where you could test a patch? Reproduction is a bit of a problem. The problem did not even occur in my own setup, but some colleagues in the SMILA team seemed to get it quite regularly. I'll try to set something up tomorrow. Thanks, Jürgen.
