[Sugar-devel] Current status of collaboration work

[James Cameron]

Thanks all for the thread and replies.

I'm not sure I understand the situation fully yet, but I'll make some
comments regardless.  Hopefully any disconnect between my comments and
your understanding will help fix mine.

I've learned long ago, to not change network protocols in a way that
breaks things.  Interoperability is critical, regardless of difficulty
of implementation or complexity.

The Tubes API is effectively a network protocol.

If we release a version of Sugar that is incompatible with previous
versions of Sugar, at a network protocol level, what will happen?

If we release a version of an activity with incompatibility, what
will happen?

We've seen what happens, even with the trivial problems introduced by
the activity.info file and the Gtk3 port.  The bias of the first
failure report is severe and lasting.  Our users don't upgrade.  They
persist with an old version of Sugar, like 0.96 on OLPC OS 12.1.0.  In
effect, they ignore the Sugar developer community until the job is
finished, the problems are fixed, or sufficient reasons are built up
to upgrade.  During which we have no end-user feedback into the
development process.  It has taken years to recover from the post-0.96
breakage.

Tony made a remark about Fedora amok.  Perhaps it was deeper than
that, perhaps it was a failure of advocacy for the Tubes API, in the
context of the Telepathy community, by members of the Sugar Labs
developer community.  Rhetorical; did they know we were using it?

If it was a failure of advocacy, we must expect further instances.
Some other API will disappear.

I've seen no detailed analysis of adopting the Tubes API into Sugar.
Why not?  What is so hard about taking the latest version of Tubes API
and integrating it into the Sugar code base in a way that we can
continue to use it?

Pull request 282 mentions 0.17.25, but of what package?  When did the
Tubes API get removed from the Telepathy packages?  Yes, for Fedora it
was 22, but we care about other downstreams.

On Fedora 18 builds with Sugar 0.107.0, I'm using

telepathy-mission-control 5.14.0
telepathy-salut 0.8.1
telepathy-gabble 0.16.7
python-telepathy 0.15.19
telepathy-glib 0.20.4

On Fedora 20 builds, I'm using

telepathy-mission-control 5.16.3
telepathy-salut 0.8.1
telepathy-gabble 0.18.2
python-telepathy 0.15.19
telepathy-glib 0.22.0

On Ubuntu 14.04 builds with Sugar 0.107.0, I'm using

telepathy-mission-control 5.16.1
telepathy-salut 0.8.1
telepathy-gabble-legacy 0.16.7
python-telepathy 0.15.19
libtelepathy-glib0 0.22.1

On Ubuntu 15.10 builds with Sugar 0.107.0, I'm using

telepathy-mission-control 5.16.3
telepathy-salut 0.8.1
telepathy-gabble-legacy 0.16.7
python-telepathy 0.15.19
libtelepathy-glib0 0.24.1

On Fedora 24 koji.fedoraproject.org says we'll expect

telepathy-mission-control 5.16.3
telepathy-salut 0.8.1
telepathy-gabble 0.18.2
python-telepathy 0.15.19
telepathy-glib 0.24.1

I don't have a test case for collaboration failure due to Tubes API
missing.  The feature page doesn't say.  I can't find a bug report.

The constant CHANNEL_TYPE_TUBES is to be removed?

I've seen patches that remove it from

sugar-toolkit-gtk3:src/sugar3/presence/activity.py

But none that remove it from

sugar-toolkit:src/sugar/presence/activity.py

Which is still used by older activities, right?

On the activity set bundled with OLPC OS 13.2.6, references to
CHANNEL_TYPE_TUBES can be found in activities; Record, Browse, Read,
ImageViewer, Pippy, StopWatch, TurtleBlocks, TurtleBlocks, Distance,
Calculate, Physics, Memorize, and Write.

The feature page doesn't list these or other activities to be
changed.  Do we know, or are we just hoping to remember them all?

I'm also wanting to see an interoperability matrix of some sort; will
new activities work on old Sugar, will old activities work on new
Sugar, etc.  I'll need to take that into account for users upgrading
laptops, where either Sugar or activities may be upgraded independently.

It would seem that the Tubes API was fairly critical to Sugar's
success.  ;-)

On Fri, Dec 25, 2015 at 08:55:18PM -0300, Gonzalo Odiard wrote:
> Hi Martin,
> I like your proposal of use the wrapper in the activities by at least one
> cycle, before include it in the toolkit.
> In our experience, once the code is included in the toolkit, is difficult make
> changes without breaking activities in
> unexpected ways.
> I didn't have time to make tests with the wrapper, and is really difficult do
> tests for collaboration. We have seen
> bugs that appear only when you have many computers, or using jabber but not
> when using the mesh, etc.
> I think the wrapper is a very, very good start (Thanks Sam and Walter) and even
> they provided patches for some activities.
> Sadly, some of the activities are on my hands, but I didn't have time the last
> months to do the proper testing
> and integration of the patches.
> About the wrapper API, just looking at the code, I think would be better add a
> callback parameter to the setup() method
> because the initialization is async and then is the only way to execute your
> activity code when the initialization
> has finished. Issues like this are difficult to get right at the first time.
> I know I am not doing almost any work in sugar these months, don't take these
> comments as a critic,
> just as a way to try to help, and avoid problems in the future.
> Regards,
> 
> Gonzalo
>    
> 
> On Wed, Dec 23, 2015 at 4:52 PM, Martin Abente <[1]
> martin.abente.lahaye at gmail.com> wrote:
> 
>     Hello everyone,
> 
>     I have been reviewing the current state of the collaboration proposals and
>     I am afraid it is still too early for merging it. We need to explore more
>     use cases, and this will only happens when we start porting more Activities
>     that actually use TUBES. Therefore, i want to share some thoughts on this.
> 
>     Opinions:
>      1. There haven't been enough changes in the Activities regarding Tubes
>         deprecation.
>      2. Dropping the Tubes support from Sugar without changing all the
>         activities that depend on Tubes means that we will break collaboration
>         for those activities anyway, and there wont be much gain by just doing
>         that.
>      3. Making changes in the Sugar API without proper testing with more
>         activities (and scenarios) is simply not a good idea.
>      4. But, making changes in the Activities can be easily handled since they
>         are self contained.
>      5. Most of our users still use Fedora 18 through OLPC deployments, where
>         Tubes is available.
>     Suggestions:
>      1. Lets make Sugar handle the Tubes deprecation better so it doesn't
>         break, but lets not remove the support for TUBES yet.
>      2. Instead, we can start changing the activities using the Wrapper that
>         Walter and Sam prepared, but using it locally on each Activity for now.
>      3. Once and if, we have most of our activities ported to the new telepathy
>         API (which will be based on the Wrapper), then we can include the
>         Wrapper into sugar-toolkit-gtk3, in a next release and remove it from
>         Activities.
>     Pros:
>      1. We avoid breaking collaboration for (a) Activities that use TUBES and
>         run on older systems where TUBES is available, and (b) Activities that
>         does not use TUBES on newer systems where TUBES is no longer available.
>         This _is_ an improvement versus the current situation where is
>         completely broken on newer systems.
>      2. We do this whole re-work incrementally, without having to change the
>         API (sort of) blindly.
>      3. There will be more flexibility to explore ideas in Activities land.
>     Cons:
>      1. There will be repeated code in Activities, but that can be changed
>         easily later.
> 
>     What would be needed:
>      1. To detect if there is TUBES support, as Sam mentioned in his first PR
>         [1]. Can someone look into this?
>      2. Do not create TUBES channel when there is not support. This [2] is just
>         a hack and the logic works fine, but it depends on whether or not we
>         can detect support.
>      3. Cleanup the Wrapper and make sure that it is possible to use it locally
>         in activities.
>     Other improvements that we could land now:
>      1. Give more flexibility to activities to use file transfer channels
>         without having the shell messing with them. [3]
>     Conclusions:
> 
>     If we don't do something about this, next Sugar releases will still be
>     broken for collaboration, for more scenarios than necessary.
> 
>     Let me know what you guys think,
>     Martin.
> 
>     Refs:
>     [1] [2]https://github.com/sugarlabs/sugar-toolkit-gtk3/pull/270
>     [2] [3]https://github.com/tchx84/sugar-toolkit-gtk3/commit/
>     bed0ac5f4259ff1669323db26acb27f5d9c8ed1f
>     [3] [4]https://github.com/sugarlabs/sugar/pull/621
> 
> --
> Gonzalo Odiard
> 
> SugarLabs - Software [for | by] children learning 
> 
> References:
> 
> [1] mailto:martin.abente.lahaye at gmail.com
> [2] https://github.com/sugarlabs/sugar-toolkit-gtk3/pull/270
> [3] https://github.com/tchx84/sugar-toolkit-gtk3/commit/bed0ac5f4259ff1669323db26acb27f5d9c8ed1f
> [4] https://github.com/sugarlabs/sugar/pull/621

-- 
James Cameron
http://quozl.netrek.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: avoid-the-tubes-api.png
Type: image/png
Size: 195524 bytes
Desc: not available
URL: <http://lists.sugarlabs.org/archive/sugar-devel/attachments/20151226/531b6eff/attachment-0001.png>