Tested the case where user follows a link while a long XML
document is loading.
The document continued loading until the new one started to
load, which (I am told) is not the wanted behavior.
Fixed the above problem by hoisting DummyParserRequest
to nsContentSink and using the DummyParserRequest
on the XML side.
Uploaded a new patch. The new patch addresses all know issues
with the previous patches (except for problems that aren’t
regressions but existed even without the patch—like stopping
scripts on Yellow Screen of Death).