[Barry-devel] Updated Barry roadmap

[Chris F. <cd...@fo...>]

Hi folks,

I've updated the TODO file in CVS, and will post it below as well.
It includes some of the desired features that we still hope to implement
in Barry, and areas of interest that other developers can help with.

If you have comments or suggestions or feature requests or questions
about some of the work detailed below, please respond to this thread.

If I've missed your favourite feature, please be sure to let me know
as well. :-)

I'm more than happy to help anyone who is interested in tackling these
tasks before I get to them.  The better Barry works for you, the better
it works for everyone.

Thanks, and I look forward to your comments,
- Chris



Porting opensync plugin to opensync 0.40
----------------------------------------
There's two options to this item:

	- simple way
	- proper way

The difference between opensync 0.22 and 0.40 involves some API changes.
You should be able to note the changes in the example plugin code once
0.40 is released.

The simple way involves merely updating Barry's opensync plugin to match
the new API.  This shouldn't be too difficult, but you may run across
some small surprises.

The proper way involves:

	- switching from the vcard/vevent formats to the opensync XML
		formats
	- switching from storing state information in text files to
		storing it in the built in database (optional)
	- making use of the new opensync time APIs to properly support
		timezones for all time operations

Switching away from text based vcard and vevent formats will remove
the burden of raw data parsing and formatting from the plugin itself,
and make use of the more tested opensync library.  Any bugs fixed
in opensync's parsers will automatically fix bugs in the Barry plugin.

Switching state storage formats may allow for greater flexibility
in supporting multiple devices.  This needs more research, but it
is the "way things are done" in opensync, and likely worth moving
in that direction.

Support for timezones will likely stress the opensync API as well as
the Barry API, but definitely needs to be done for completeness on both
sides of the equation.

Estimated time:   simple way: 10 hours if lucky
                  proper way: open ended



Reverse engineer the new TimeZones database
-------------------------------------------
This is a relatively light but important project for those that want to
get their feet wet.

In relation to the above opensync porting, the new TimeZones database
found in devices such as the BlackBerry 8830, will need to be reverse
engineered in order to fully support timezone conversions.

I have sample data of these databases in hex dump format for anyone
who wishes to take a stab at it.  If you have a newer device,
you can retrieve this data yourself using:

	btool -d "Time Zones"

Estimated tasks:
	- reverse engineer the record
	- write and test the parser
	- no builder needed, as it is a read-only database
Estimated time: 4 hours



Reverse engineering java loader protocol
----------------------------------------
This has not been done by any opensource Blackberry project out there,
to my knowledge, and would be most useful to Blackberry application
developers.

If you are a Blackberry app developer, this may interest you.

Estimated time: open ended



Adding GPRS / CDMA modem support
--------------------------------
Barry has had plans to include modem support for a few months now, but
it has not yet percolated to the surface.

The primary source for modem protocol specifics is the XmBlackBerry
project (http://sourceforge.net/projects/XmBlackBerry).  Barry's
goal is to create a standalone pppob utility that can be configured
with pppd similarly to pppoe.

The tricky part of modem support is that it requires threading support,
especially if you intend to use the database at the same time.
This leads to some architectural challenges that need to be dealt with:

	As Barry is not a single application, how do you access the
	database while pppob is using the modem?

There are two viable ways of dealing with this.  One involves placing
a (hopefully thin) driver in the kernel, and the other involves using
a daemon and RPC calls.

My preference is to implement this using RPC calls if needed, and hammer
out all the implementation details in user space.  Once they are well
understood, a smaller kernel driver hook may be more easily written
that supports routing messages according to socket or application
needs.  For example, one application may register an interest in
database messages, another in javaloader messages, and another in the
multiple modem socket messages.

There is currently a development branch in CVS that intends to support
optional threading support in the Barry library, so that at least
it should be possible to use both the database and the modem in the
same application.

Estimated tasks (assuming threading support is finished):
	- copy protocol specifics from XmBlackBerry
	- add Serial class for controller interface
	- write and test pppob
	- write pppob manpage
Estimated time: 8 hours



Finish the threaded architecture changes in CVS branch
------------------------------------------------------
For the reasons outlined above, for modem support, this CVS branch
needs to be finished.  For more information, contact the mailing list.

Estimated tasks:
	- write SocketRouting code
	- refactor the controller class into RAII behaving
		mode classes, such as Desktop, Serial, and
		Javaloader, that operate on top of Controller
	- testing threaded and non-threaded behaviour
	- testing performance impact
Estimated time: 20 hours



Add bluetooth serial support
----------------------------
It is reported that it is possible to access the database through
Bluetooth using the older Blackberry serial protocol.  XmBlackBerry
has support for this and may be used as a reference.

The goal here would be to hide the bluetooth access behind the
same Barry library API, so that syncing with the opensync plugin
would be seamless whether plugged in via USB or Bluetooth.

Estimated tasks:
	- research and design serial protocol stack to reuse as much
		library code as possible
Estimated time: unknown



Write simple GUI for streamlining sync setup and action
-------------------------------------------------------
Syncing setup and operation is currently a tedious, complicated task.
A GUI that performed all the detailed setup and configuration work,
for a Blackberry-specific sync, using opensync libraries and plugins
for Evolution, Sunbird, etc, would be very helpful.

This would be much easier for an experienced GUI programmer, but there is
a learning curve for the opensync API.

Estimated tasks:
	- document the settings required for Blackberry, and
		all intended plugins required
	- write application that:
		- does the opensync configuration through the opensync
			API directly
		- scans the USB bus for available Blackberry devices using
			Barry
		- lives in the system tray watching for Blackberry devices
	- if aiming for super ease of use, script a source build of
		all needed opensync components and install in private
		area to avoid conflict with system
Estimated time: unknown
Note: Depending how fast HAL, OpenSync, and Conduit are implemented,
	this may never be needed... but if it existed today, there's a
	lot of users who would be very happy...



Document the USB protocol
-------------------------
Currently the only english documentation for the Blackberry protocol
is the webpage by the Cassis project (found at
http://off.net/cassis/protocol-description.html).

The USB protocol is not nearly so well documented.  The best documentation
available can be found in the Barry header files: protocol.h and
protostructs.h.

Translating the code into documentation (into a wiki, that will hopefully
soon be available) is a great way to get involved in the project and
learn a lot about the Blackberry from a low level.

Unfortunately, Jedi mind tricks don't often work when trying to convince
people to write documentation for me... :-)

Estimated tasks:
	- write, write, write
Estimated time: 40 hours (documentation expands to fill available time...)



Code cleanup
------------
Code can always be improved.  There are two big ways to help:

	- write an application using the Barry library
	- improve the Barry library itself and send patches

By writing an application, you can provide crucial feedback on the ease
of use of the Barry API.  I'm very eager for such feedback.

Secondly, there is currently a lot of code ducplication in the record
classes, and a careful refactoring is required.  I would be open to a class
hierarchy, with possibly private or protected inheritance.  My primary
concern is object safety when using the record classes as objects in
STL containers, with a secondary concern to make it easier to
abstractly work with a record.  This implies a careful mix of
virtual functions and a generic record base class.  Patches in this
area will be thoughtfully considered.

Estimated tasks (refactoring):
	- design safe hierarchy
	- move common code to base class
	- make sure all record classes use the common record API
	- test
Estimated time: 7 hours



C API wrapper
-------------
For those that write applications in C, a C API wrapper has been started
in the cbarry.h header.  It has not yet been implemented, but should be
straightforward.

Estimated tasks:
	- finish some API design work (head not fully complete)
	- implement all functions (about 50)
	- write test application, or test suite, for C API
Estimated time: unknown



Python wrappers and example code
--------------------------------
For those that write applications in Python, a SWIG wrapper has been
started by H Miz Jones.  This is partially functional, and involves
working with the Barry API, and may introduce changes to it depending
how hard it is to translate things to the Python world.

The SWIG wrapper scripts have not yet been publically released, but
please contact me if you are interested.

Estimated tasks:
	- finish C++ / Python integration (possible template issues)
	- finish SWIG wrapper
Estimated time: unknown



Command line backup and restore
-------------------------------
The only command line backup currently available is the one in btool,
using the -f and -s switches.  This does not backup exact data from
the device, but parses it, stores it in the Boost serialization format,
and then reverses the process for restore.  This is a great test
for the Barry library, but not so great for backup, since not all
databases can be parsed.

There is already an exact backup and restore interface with the GUI, but
there is a lot of useful functionality trapped in a layer of GUI
that could be just as useful from the command line.  Tasks such as a nightly
cron backup of any Blackberry devices attached to the system would be more
easily done via command line.

You could add command line arguments to the barrybackup program to skip
the GUI (tricky and possibly error prone), or you could pull the backup
functionality into a standalone command line utility (more work, but smarter
in the long run).  This is mostly a code refactoring job, consisting of
all working code that's already there, and I know there are people
who would thank you. :-)

Estimated tasks:
	- split out tar and backup functionality code into shared library
	- write and test command line tool
Estimated time: 6 hours