[Opensync-users] Is Syncing Right?

[Martin O. <doc...@gm...>]

Dear all,

I'm been thinking for a few weeks about pim support in linux. That and
how various models for accessing data from devices, online services and
local stores.

My conclusion is that a syncing process (the idea as it stands in
OpenSync) is the wrong approach fundamentally. We're caught up thinking
about conduits, getting data from one place to another, converting it
between all the weird and wonderful formats out there.

But I suggest, there is a different way of thinking about data access.
Instead of thinking about transport. First we should have a look at
categorical access to the data; what I mean by this is that we have
various data categories (addressbook/photo album/music playlist/ect) and
that providing a standard api to each of these categories without
restricting anything to any specific backend or client, would allow us
to add data services on the fly. Much like the way music is handled when
you plug in an ipod and libgpod together with hal via dbus tells
rythembox or amorock that there is a device that contains music
available. The user then accesses and plays the msuic directly fromt he
device, or at their option can then sync or copy data between local
stores and the device.

The same would be said of online services: When you specify your google
information, the system should show your calendars in the calendar data
services; but only if your online. Since when offline it's logically
equiverlent to being unplugged from a device.

So, I suggest these services run over dbus and be independent enough to
allow even cli tools to be based on them.

Once you have data access, automatic syncing, caching and a whole host
of other interesting features becomes possible.

But until we start thinking more about our data stores and user data
services more seriously, I think we'll continue to see the hacks and
copies of windows ideas about how this sort of thing should work.

Regards, Martin