Index: /trunk/archive/pslib/testing/pap_rel2.txt
===================================================================
--- /trunk/archive/pslib/testing/pap_rel2.txt	(revision 1683)
+++ /trunk/archive/pslib/testing/pap_rel2.txt	(revision 1684)
@@ -1931,2 +1931,38 @@
 L1720 minNum, maxNum
 L1744 Non-uniform histogram bins
+
+=============================================================================================================
+
+To get psLogMsg output to stderr, you need to specify:
+        psLogSetDestination("dest:stderr");
+Similarly for stdout.  I can see where they got it from (the SDRS
+entry could easily be misinterpreted), but I don't think this is the
+behaviour we expected.
+
+psTrace:
+
+* Each "facility" must commence with a dot: e.g.,
+".IPP.debias.overscan", which I think is really annoying because it's
+easy to forget.
+
+* You must set every level for each node on the trace tree explicitly if
+you intend it to be used.  For example, the call
+        psTraceSetLevel(".IPP.debias.overscan", 5)
+will produce the following setup:
+.                    0
+ IPP                 5
+ debias              5
+ overscan            5
+Note how each of those is in the same level, so that the hierarchy
+hasn't been built up?  Of course, this is a bug because it does
+something completely unexpected, but I would appreciate confirmation
+of my expectation that a 4-deep heirarchy (".", ".A", ".A.B",
+".A.B.C") should be formed from simply specifying a level for
+".A.B.C".
+
+* I would have thought that the property of inheritance that we
+specified implied that if I hadn't explicitly set the level for
+".A.B", but I have set the level for ".A", then printing a trace
+message on component ".A.B" would depend on comparing the level to the
+set level of ".A"; but in the implementation, it would be ignored
+completely.