Shared writable blob directory causes various errors if flag is not set

[Christian Theune]

If the ZEO server and ZEO client both use the same directory (aka shared writable) and the flag is not set,
we can run into nasty problems that aren't obvious to understand.

Some of the effects include:

10:16 < zagy> 2007-06-22T10:16:24 CRITICAL txn.30661632 A storage error
              occurred during the second phase of the two-phase commit.
              Resources may be in an inconsistent state.

"/Users/zagy/development/eggs/tmp877SSZ/ZODB3-3.8.0b1-py2.4-macosx-10.4-i386.egg/ZEO/TransactionBuffer.py", line 122, in clear
09:57 < zagy> ValueError: need more than 2 values to unpack

IMHO we should be able to discover this state (a shared writable directory) and either fail if that doesn't fit the configuration parameters or automatically assume a shared writable directory.

[Christian Theune (ctheune)]

We can fix this by making shared blob directories automatically discoverable. However, this is not a bug but a feature that can be implemented differently without being so error prone.

Therefore we're not gonna change this for ZODB 3.8 but add a new feature in the future, hopefully in ZODB 3.9.