Collaborative Activity Development

[Dafydd Harries]

Ar 14/04/2008 am 13:02, ysgrifennodd John Gilmore:
> "Morgan Collett" <morgan.collett at gmail.com> wrote:
> > I'm now working for OLPC, on improving activity collaboration.
> 
> Tying collaboration to Sugar is a losing strategy.  Once the rest of
> the world figures out that *their* programs should be trivial to
> collaborate in too, they'll reimplement collaboration (likely in an
> incompatible way).  Then Sugar's collaboration will be an orphan
> rather than the mainstream.  Instead, if OLPC's collaboration code
> supported cross-platform collaboration, OLPC's model and its
> implementing code would spread throughout the whole computing
> infrastructure.  And that would bring in a new pile of contributors,
> enhancing, debugging, and porting it everywhere.

I absolutely agree. While most of the OLPC collaboration code will happily
work outside Sugar, there are interoperability problems, and I'm not sure what
we should do about some of them. I think perhaps we made a mistake in focusing
on designing a good API rather than protocols that would be general and easy
to reimplement (specifically, I think the discovery protocols are maybe not
general enough, and the data transport protocols are maybe too hard to
reimplement).

That said, I and a number of other people are working on popularising the
collaboration software that Collabora developed for Sugar outside Sugar.

> PS:  If there is a simple way to install a couple of RPM's or DEB's, add
> a paragraph of code and a few automake macros, and add collaboration
> to any program written in C or C++, then please document it!  (If on
> the other hand "it only works in Python" and "requires sugar-jhbuild"
> then there's some work to be done.)

It's not limited to Python, and doesn't require sugar-jhbuild, but our APIs
are too complicated. I'm hoping that we can adopt things in the vein of Ben
Schwartz's work to provide nice ways for developers to make their applications
collaborative.

Ideas about how to make our collaboration technology more general and
accessible are welcome.

> 
> PPS:  This review of AbiWord says:
> 
>   http://www.linux.com/feature/131852
> 
>   The new AbiWord supposedly offers real-time document collaboration
>   developed for the OLPC project and implemented by means of an
>   experimental plugin. As per the AbiWord-2.6 release notes, there are
>   three implementations of the plugin, one for the OLPC, and two (an
>   XMPP-based one and a pure TCP/IP one) for Linux. The Linux plugins
>   compiled without any issues, but AbiWord couldn't activate them. The
>   plugin isn't currently available for Windows.
> 
>   [abiword.com says the Windows plugin is available on 2.6.2 now.  But
>   I never did find the collab plugin, nor any documentation for it.  The
>   2.6.0 release notes imply that the three available collab plugins can't
>   actually interoperate with each other!]

I don't know of any reason why the OLPC plugin shouldn't work on a Linux
system, barring perhaps some dependency problems (e.g. new enough version of
libdbus).

-- 
Dafydd