The PutField API appears to provide 2 ways of writing an object's field values
to a serialization stream. The recommended method is to call
ObjectOutputStream.writeFields() after setting values in the PutField object:

    private void writeObject(ObjectOutputStream out) throws IOException {
        ObjectOutputStream.PutField fields = out.putFields();
        // set field values ...
        out.writeFields();
    }

However, ObjectOutputStream.PutField also provides a public write() method,
which accepts an object implementing ObjectOutput as a parameter. This presents
the following problems:

    1. It's unclear why there are two different methods of flushing a PutField
    to an ObjectOutputStream (javadoc/spec doesn't elaborate).

    2. PutField.write() allows the field values to be written to an
    ObjectOutputStream other than the one in which the object is being
    serialized.

    3. There is no corresponding GetField.read().

    4. Unlike ObjectOutputStream.writeFields(), PutField.write() doesn't unset
    block data mode before writing field values, nor does it reset block data
    mode afterwards. As a result, there is no way for an ObjectInputStream to
    make any use of the values written by PutField.write()
    (ObjectInputStream.readFields() expects the fields to be written in
    "non-block-data" mode).

Attached code demonstrates #4.

The implementation of PutField.write() could be changed to unset block data mode
on the given stream before writing field values, and then (re)set it afterwards.
Since PutField.write() accepts ObjectOutput (not ObjectOutputStream), it must
first explicitly check to see if the stream it's passed is an
ObjectOutputStream, and then toggle block data mode.

A better (IMO) solution is to deprecate PutField.write(), and possibly change
its implementation to throw an exception, since it never worked in the first
place.