loading index data.  This leverages the new fsBucket toString and
fromString methods and provides much faster FileStorage index saving and loading
and smaller index files.  On my machine, saves are 5 times faster and
loads are 20 times faster (after a save, when data are in disk
cache).  Indexes are roughly 30% smaller.

The index format has changed.  Old indexes can be read just fine, but
new indexes won't be readable by older versions of ZODB.

alternative to __getstate__ and __setstate__.

Also made it possible to add extra bucket methods.

Changed ZEO undo protocol. (Undo is disabled with older clients.)
Now use one-way undoa.  Undone oids are now returned by (tpc_)vote for
ZEO. Undo no-longer gets commit lock.

including those created via ``multiprocessing.Process`` instances.

This entailed giving each client storage it's own networking thread.

to be fixed. :/

- Most server methods now return data to clients more quickly by writing to
  client sockets immediately, rather than waiting for the asyncore
  select loop to get around to it.

- More clearly define client and server responsibilities. Machinery
  needed for just clients or just servers has been moved to the
  corresponding connection subclasses.

- Degeneralized "flags" argument to many methods. There's just one
  async flag.

Allow thunks to take args.

Don't hold trigger lock when calling thunks.

undoMultiple. This gives applications that want to support older
versions of ZODB a method to test for.

- The undo implementation was incorrect in ways that could cause
  subtle missbehaviors.

API changes:

- The API for undoing multiple transactions has changed.  To undo
  multiple transactions in a single transaction, pass pass a list of
  transaction identifiers to a database's undo method. Calling a
  database's undo method multiple times in the same transaction now
  raises an exception.

- The storage API (IStorage) has been tightened. Now, storages should
  raise a StorageTransactionError when invalid transactions are passed
  to tpc_begin, tpc_vote, or tpc_finish.

broken objects.

Python 2.6, where asyncore.file_wrapper was changed to dup it's the
file-descriptor passed to it. :(  Fixed the fix to work withor or not
the fd is duped.

breaking the client loop.

Cleaned up trailing whitespace.

hard to turn back on when it's needed, although, at that poiint, it
still might not be enough. :)

Object state management wasn't dome correctly when classes
  implemented custom _p_deavtivate methods.
  (https://bugs.launchpad.net/zodb/+bug/185066)

See https://bugs.launchpad.net/zodb/+bug/459889

data.  (This sometimes caused an AssertionError in
Connection._setstate_noncurrent.)

passed a tid older than anything in changes but newer than
anything in base.

test into an automated test, and included Chris Wither's repozo fix
(as monified by Godefroid Chapelle) to avoid a deprecation warning.

Bug fixed:
- Internal ZEO errors were logged at the INFO level, rather
  than at the error level.

- zope.testing was an unnecessary non-testing dependency.

Bug Fixed:

History support was broken when using stprages that work with ZODB
  3.8 and 3.9

of md5 support to avoid a deprecation warning.  I would love to
reapply 105888 (from the 3.9 branch) when there is an adequate test.
This can be achived through a pretty straightforward conversion of the
existing manual repozo test to an automated test.

Bugs Fixed:
  2 BTree bugs, introduced by a bug fix in 3.9.0c2, sometimes caused
  deletion of keys to be improperly handled, resulting in data being
  available via iteraation but not item access.