> We just released the whole PyF 2.0 stack.

> A big change in this release is that we decided to synchronize the
> releases of all the layers of the framework:
> - pyf.dataflow
> - pyf.manager
> - pyf.transport
> - pyf.splitter
> - pyf.warehouse
> - pyf.station
> - pyf.componentized
> - pyf.services

> With this change we hope to be able to more effectively keep the API
> changes in sync along the stack, as well as make it more clear which
> versions of which components are compatible. As a rule, we chose to
> have the first 2 numbers of each release to be synchronized, and use
> an eventual third number for bugfix releases, just like Gnome does for
> example. Plugins don't follow this convention. For now, we want to be
> more lax with them and let them evolve at their own pace. We also feel
> that not pressuring plugin developers to follow a strict release
> process might encourage third parties to create more plugins, making
> for a more lively PyF ecosystem.

> This is also the first release to include a stable documentation.
> Until now, the documentation was closer to the development trunk than
> the actual stable releases. The documentation for 2.0 is available athttp://pyfproject.org/documentation

> We also added a bunch of unit tests and fixed all those that were
> failing. PyF is stil far from 100% coverage, but this is nevertheless
> a first step in the right direction. In a framework composed of
> several layers, keeping the APIs in sync is of the utmost importance,
> and unit tests are possibly the only way to achieve that.

> PyF now uses the setuptools "extras_require" directive to allow
> installing different levels of PyF, along with the appropriate
> dependencies:
> - if you want the thrill of doing everything on the low level:
>     $ easy_install pyf
>   or
>     $ easy_install pyf[minimal]
> - if you like the ability to use XML networks with default plugins:
>     $ easy_install pyf[components]
> - and if you need a full-blown web application, tube designer along
> with all default plugins:
>     $ easy_install pyf[fullstack]

> If you installed pyf from the bootstrap script, updating is simple,
> just type this in your virtualenv:
>     $ easy_install -UZ pyf[fullstack]

> We now support multiprocessing: each node of the network can be in its
> own process. Of course, this introduces some message passing overhead
> and as such it is not always more efficient, so a judgement call must
> be made for each application on a case-by-case basis. Note that this
> is probably still a bit rough, and will no doubt be refined over the
> next releases. To easily test this, you will find a new "advanced"
> expandable panel on your components in the tube designer, expand it
> and select the "separate process" checkbox.

> Components saw a rather drastic change: through a metaclass, they now
> all inherit the configuration of their parent components. The same
> metaclass also automatically documents the configuration keys of the
> components. Yay for more docs!

> PyF.Services is now based on TurboGears 2.1 instead of the alpha/beta
> as previously. We strongly recommend that you migrate your
> PyF.Services applications to TurboGears 2.1. Migration should be
> pretty straight-forward, don't hesitate to ask for help on the mailing-
> list or on IRC.

> Tube parameters (enclosed in the <params> tag) were drastically
> improved. We now support more params types such as boolean, select and
> radio buttons. Multi-select is on its way, but it requires some
> invasive changes to the core of PyF.Componentized and as such we
> preferred delaying it for next release. Params can also be visually
> grouped in fieldsets. This allows making the launcher form more
> visually appealing, but it doesn't change the params behavior. We
> realize that as of now this is still an undocumented feature and we'll
> make sure to properly document it. If you want to see it done sooner,
> patches are obviously welcome. ;)

> We also fixed a lot of bugs both in the stack and in existing plugins,
> but let's not make this email even longer. If you want all the gory
> details, go read the Mercurial logs for each module. :)

> Best regards, and happy coding,
> The PyF Team !