Darren Duncan wrote:How much trouble would it be to actually look inside each file?Lots of trouble. Since DBD::DBM supports many DBM types and many MLDBM types, one file might be SDBM, another Berkeley_DB, etc. and each of those might use MLDBM or not and MLDBM can use a number of serialization methods (Data::Dumper, Storable, User-Defined,etc.). And suppose one of the serialized fields contains perl code and the serializer evals it on unserailize ... could get very ugly.
Patches are welcome if you think it's doable and worthwile :-).
In the mean-time, would it be worthwhile to change the data_sources() implementations to simply return an empty list, as 'sqlite' currently does? In my mind, that would seem preferable to returning a list of invalid items. -- Darren Duncan
