--- /dev/null
+$Id: Notes,v 1.1 2006-05-31 12:04:19 mike Exp $
+
+Some things that it would be nice to check (varying priorities!)
+
+- Response to a second Z39.50 Init
+
+- Whether referenceId is properly implemented
+
+- Whether attribute-set is inspected and honoured
+
+- How SRU servers respond to bad HTTP headers
+
+- Malformed queries such as "@and foo"
+
+Add plugin architecture for additional tests. Allow each test to
+specify whether it is CRUCIAL, DESIRABLE or whatever, and the
+frequency with which it is run.
+
+Note results of each individual test (where?) and raise an alert when
+a test gives a different result from the previous time.
+
+Capture more information about what goes wrong at low levels. Does
+YAZ itself allow an APDU log to be generated or is it still only
+yaz-client and/or GFS that does this?
+
+Make a queue of servers to be queried, and keep the pod "sufficiently
+full", throwing a new one in as an older one is exhausted.
+
+Two weeks to have framework running.
+
+Test dependent on other test; use results from previous test as input
+for this one. One test that is the aggregate of others.
+
+Have a pre-cooked set of sample words to use in queries in the hope of
+finding one or more records; remember which words work well for each
+server.
+
+Make this into Alvis work by adding a subject-sniffer for the peer
+network. Use the extended ZeeRex records to produce Alvis D3.1
+peer-description records using XSLT that Marc will write.
+