The serve-stale system test should be extended to:

1. Check that if "serve-stale-enabled" is explicitly set to "no" in the configuration file, the serving of stale records is disabled. (At present, the tests check 'serve-stale-enabled yes" and the default when serve-stale-enabled is not specified).

2. Check that if a record is stale, then before the stale record is returned, upstream queries are made.

3. Check that the TTL of the record returned is that set by the stale-answer-ttl parameter.  (This should check that the stale-answer-ttl value is used regardless of whether the TTL of the record is less than greater than this value).

4. Check that the max-stale-ttl value does cap the amount of time for which the stale record is served.

5. Check that if a record is stale (and is served) then if the authoritative server subsequently comes back but the record is changed, the new value is returned. This should check that an appropriate response is returned for:

a) A different RDATA
b) An NXDOMAIN response from the server
c) A NODATA response from the server