client might not properly reconnect to a server.  Verify
that caches are not unnecessarily flushed.

Handle expiration times

Handle READDIRPLUS

Respect opaque directories and easily specify which are opaque.

fragment when NFS block size is 8KB but the SFSRO is 16KB?

Test pulldb, make sure easy to use.

Allow reconnection to any SFSRO server without losing cache contents. 

make the sfsrodb code cleaner by consolidating code to
store big directories with file data.  indirect data pointer code.
the sleepycat database is way too big.