sync engine development ideas

[Mark Ellis]

Hi Guys (John, copied you in in case your working on anything
significant).

I've been meaning to do a few things to sync engine, and I've finally
got around to figuring out how I want some of it to work. This is a
request for comments, ideas, help with some sticking points, and sanity
checking :)

It started with wanting to get pre WM5 devices into sync engine, at
which point I decided this would be easier if sync engine was designed
to handle multiple devices, because then you just support 2 classes of
device. Of course multiple devices would be cool anyway, so I'll do that
first.


The first bit is quite easy, move a lot of stuff from the SyncEngine
class into a new Device class, and have SyncEngine maintain a collection
of Devices. I intend to then have an 'active' device which all sync
activities act on. This is not complete multi device functionality, but
would allow multiple devices to connect without confusing sync engine,
which I believe would cause problems at the moment.


The next step I haven't completely ironed out yet, but would be
something like this.

Each Device would advertise the dbus interface on its own object path,
either using the device id or partnership id as part of the path, I'm
not sure which. The opensync sync group would then have device and
partnership info as part of its config, so it could contact the correct
dbus object. This would also tie an opensync group to a partnership,
which is definitely a good thing.

I've currently got two problems with this second part. Firstly, creating
a partnership and sync group would have to be combined for everyday
users, and I can't decide the best way to do that.

The next problem is a little more complex. To sync multiple devices
needs multiple sync groups, but we've gone a way to doing that above
with different configs. However, to enable device initiated sync, we
need to know the name of the sync group for that device to run the
correct command. I guess this could just be in the partnership specific
config, but I don't think this is well known. Maybe some kind of config
editor ?

So, what do you guys think ?

Mark

[Guido Diepen]

Hi Mark,

Post by Mark Ellis
Hi Guys (John, copied you in in case your working on anything
significant).
I've been meaning to do a few things to sync engine, and I've finally
got around to figuring out how I want some of it to work. This is a
request for comments, ideas, help with some sticking points, and sanity
checking :)
It started with wanting to get pre WM5 devices into sync engine, at
which point I decided this would be easier if sync engine was designed
to handle multiple devices, because then you just support 2 classes of
device. Of course multiple devices would be cool anyway, so I'll do that
first.

I do like the idea of just having one sync-engine that can handle both the
post wm5 devices, as well as the pre wm5 devices. Regarding the multiple
devices, I still have not gotten around to test this, but does it actually
work in Windows to have multiple WM5/6 devices connected at the same time?

Post by Mark Ellis
The first bit is quite easy, move a lot of stuff from the SyncEngine
class into a new Device class, and have SyncEngine maintain a collection
of Devices. I intend to then have an 'active' device which all sync
activities act on. This is not complete multi device functionality, but
would allow multiple devices to connect without confusing sync engine,
which I believe would cause problems at the moment.

I have to say that this idea sounds pretty sane to me.

Post by Mark Ellis
The next step I haven't completely ironed out yet, but would be
something like this.
Each Device would advertise the dbus interface on its own object path,
either using the device id or partnership id as part of the path, I'm
not sure which. The opensync sync group would then have device and
partnership info as part of its config, so it could contact the correct
dbus object. This would also tie an opensync group to a partnership,
which is definitely a good thing.

Indeed the last part is indeed a good thing. It would also allow for
aborting cleanly in case you are using a different device with the
opensync group, instead of messing up everything.

Post by Mark Ellis
I've currently got two problems with this second part. Firstly, creating
a partnership and sync group would have to be combined for everyday
users, and I can't decide the best way to do that.

One possibility would be creating a partnership name in the opensync
configuration for our synce plugin. You could then see if it is possible
to have the opensync-synce plugin talk to sync-engine, see if the
partnership exists. If it does not exist AND if there is a free slot, the
partnership with the given name will be created automatically.

This way you would have all the work done by the opensync plugin.

Using the above would enable users to only have to set the partnership
name in the configuration for opensync, and the partnership would be
created automagically.

Let me know what you think about the above approach.

Post by Mark Ellis
The next problem is a little more complex. To sync multiple devices
needs multiple sync groups, but we've gone a way to doing that above
with different configs. However, to enable device initiated sync, we
need to know the name of the sync group for that device to run the
correct command. I guess this could just be in the partnership specific
config, but I don't think this is well known. Maybe some kind of config
editor ?

One possibility could be to have the same name for the partnership and the
opensync group to be able to use this feature?

Guido Diepen

--
Aviation is proof that given the will, we have the capacity to achieve
the impossible.
--Eddie Rickenbacker

[Dr J A Gow]

Post by Guido Diepen
Hi Mark,

Post by Mark Ellis
Hi Guys (John, copied you in in case your working on anything
significant).
I've been meaning to do a few things to sync engine, and I've finally
got around to figuring out how I want some of it to work. This is a
request for comments, ideas, help with some sticking points, and sanity
checking :)
It started with wanting to get pre WM5 devices into sync engine, at
which point I decided this would be easier if sync engine was designed
to handle multiple devices, because then you just support 2 classes of
device. Of course multiple devices would be cool anyway, so I'll do that
first.

I do like the idea of just having one sync-engine that can handle both the
post wm5 devices, as well as the pre wm5 devices. Regarding the multiple
devices, I still have not gotten around to test this, but does it actually
work in Windows to have multiple WM5/6 devices connected at the same time?

Fundamentally this is a good idea.

I basically started to re-work the syncdb component of sync-engine prior
to integrating it with the functioning components of librra to allow
pre-WM5 devices - then the only WM2002 device I have for testing decided
to stop working, so I couldn't test it (problem with the HAL component
of synce). This was some time ago, things have moved on so will try
again to get the WM2002 device connected and see if it works this time!

Post by Guido Diepen

Post by Mark Ellis
The first bit is quite easy, move a lot of stuff from the SyncEngine
class into a new Device class, and have SyncEngine maintain a collection
of Devices. I intend to then have an 'active' device which all sync
activities act on. This is not complete multi device functionality, but
would allow multiple devices to connect without confusing sync engine,
which I believe would cause problems at the moment.

I have to say that this idea sounds pretty sane to me.

Post by Mark Ellis
The next step I haven't completely ironed out yet, but would be
something like this.
Each Device would advertise the dbus interface on its own object path,
either using the device id or partnership id as part of the path, I'm
not sure which. The opensync sync group would then have device and
partnership info as part of its config, so it could contact the correct
dbus object. This would also tie an opensync group to a partnership,
which is definitely a good thing.

Indeed the last part is indeed a good thing. It would also allow for
aborting cleanly in case you are using a different device with the
opensync group, instead of messing up everything.

Post by Mark Ellis
I've currently got two problems with this second part. Firstly, creating
a partnership and sync group would have to be combined for everyday
users, and I can't decide the best way to do that.

Multiple devices would be an ideal - it should not be too difficult to
map a partnership to a specific device. The internal syncdb is
encapsulated and this should not be at all difficult on the
SynCE/sync-engine side. The problem would be on the opensync side-
enforcing the mapping between an opensync group and a sync-engine
internal syncdb.

Post by Guido Diepen
One possibility would be creating a partnership name in the opensync
configuration for our synce plugin. You could then see if it is possible
to have the opensync-synce plugin talk to sync-engine, see if the
partnership exists. If it does not exist AND if there is a free slot, the
partnership with the given name will be created automatically.

I like this approach - it would make sensible use of the opensync
configuration mechanism and would enforce Opensync group bindings
to both a specific device and a specific partnership.

The current per-partnership XML configuration could be placed in the
Opensync plugin configuration - this would suit Opensync 0.4x well.

We would need a mechanism to delete dangling partnerships on the
sync-engine side if they have been deleted by connection of the phone to
another host - otherwise they will get recreated when the phone is next
plugged in.

Post by Guido Diepen
Using the above would enable users to only have to set the partnership
name in the configuration for opensync, and the partnership would be
created automagically.
Let me know what you think about the above approach.

Would work well, I think. And enable us to eliminate the command-line
partnership creation.

John.