Recent public API changes:

- Since 0.1.2b beta release: Factory now takes EncodingFinder and
  ResponseTypeFinder class instances instead of functions (since
  closures don't play well with module pickle).

- ClientCookie has been moved into the mechanize package and is no
  longer a separate package.  The ClientCookie interface is still
  supported, but all names must be imported from module mechanize
  instead of from module ClientCookie.  Python 2.3 is now required. (I
  have no plans to merge ClientForm with mechanize.)  Note that the
  logging work-alike facility is gone, and the base logger has been
  renamed from "ClientCookie" to "mechanize".  Also, the experimental
  BSDDB support is now only included as example code, and not
  installed, and the VERSION attribute has been removed (mechanize
  still has its __version__ attribute).

- pullparser has been moved into the mechanize package and is no
  longer a separate package.  Also, the interface of pullparser is no
  longer supported.  Instead, it's just a purely internal
  implementation detail of mechanize.

- Removed mechanize.Browser.set_seekable_responses() (they're always
  seekable anyway).

- Some mechanize.Browser constructor args have been moved to
  mechanize.Factory (default_encoding, ...).

- .get_links_iter() is gone (use .links() instead).

- .forms() and .links() now both return iterators (in fact, generators),
  not sequences (not really an interface change: these were always
  documented to return iterables, but it will no doubt break some client
  code).  Use e.g. list(browser.forms()) if you want a list.

- .links no longer raises LinkNotFoundError (was accidental -- only
  .click_link() / .find_link() should raise this).

- Rename set_credentials --> set_password_manager (and add some new
  methods to improve auth and proxy support).

- Added response.get_data() and .set_data() methods, and make responses
  copy.copy()able.  Browser has a .set_response() method.  responses
  returned by the Browser are now copies, which means that other code
  altering headers and data and calling .seek() won't mess up your copy of
  a response.

- mechanize.Factory has changed completely, to make it easier to avoid
  re-parsing (principally: add .set_response() method and make
  factory methods take no args)

- mechanize.Browser.default_encoding is gone.

- mechanize.Browser.set_seekable_responses() is gone (they're always
  .seek()able).  Browser and UserAgent now both inherit from
  mechanize.UserAgentBase, and UserAgent is now there only to add the
  single method .set_seekable_responses().

- Added Browser.encoding().

- Factory() takes an i_want_broken_xhtml_support argument, as a stop
  gap until I actually make a proper job of it.  Without a true value
  for that argument, mechanize is ignorant of XML/XHTML.

- _authen handler name renamed --> _basicauth