[Standards] Could anyone using XMPP-0277 or PEP help us test our Onesocialweb implementation ?

Laurent Eschenauer laurent at eschenauer.be
Wed Apr 14 19:56:59 UTC 2010


>
> Just wanted to tell you guys that we changed the notification engine
>> of Onesocialweb from our crappy custom stuff to.. XEP-0277 ! We made
>> quite a few custom changes to support activitystreams, per item
>> privacy etc.. but still remain compatible with the overall
>> PEP/XEP-0277 protocols. I'll ping an email later with suggestions on
>> evolving PEP and XEP-0277 based on our work.
>>
>
> I think the XEP-0277 is not complete and may be buggy.
>
>
It is definitively not complete and support only limited use cases. We
customized a lot for onesocialeweb (details below) however we tried to be
'backward' compatible with the spec as it is today. In the future, we would
love to see these ideas be generalized in PEP and/or PubSub, and
definitively need the help of the community for this.



> Consider the situation with users A, B and C.
> [...]
> So it looks like something's missing, perhaps a Cc-like mechanism.
>

Correct. We use such a cc mechanism in OSW. At this stage we rely on the
<thr:reply-to/> of the atom payload. In the future, it would be nice to
decouple and makes this a broader pep/pubsub feature (ability to post to a
node while cc other jids). Maybe adding an <addresses/> field to the
<publish/> or <item/> ?


> Also, the use of the <thr:reply-to/> element can create privacy issues
> since it theoretically contains the JID of the previous poster which is sent
> to the replier's contacts.


Two modes of operation for the reply to:
a) replying to 'someone' (a bare jid), the equivalent of @replies in
twitter, or writing on a wall in facebook
b) commenting on an item (jid + node + item id)

We handle these cases differently (in theory, case b is not implemented
yet):

a) just like a regular update + cc (like described above)
b) the comment is not posted by the author to its node, instead, it is
pushed to the original item author, and he is the one redistributing the
comment based on the privacy settings of the item . When other users browser
his node, they will see the item and the comments attached to it (if
authorized of course).

^^this is short explanantion, happy to deep dive in another thread :-) The
key thing is that for us, the JID is really a public identifier (like the
fact that i'm @eschnou on twitter). What matters is to protect the data
behind that identifiers in a way that respect a user privacy choice (in the
vcard, activities, etc...).


> I think this case should be addressed, specially in the light of the Google
> Buzz privacy issues. Similarly, I'm wondering if there shouldn't be a public
> and a private roster, but it makes things much more complicated.


Yes, these cases are not straightforward and require a good balance between
flexibility, simplicity of implementation, privacy control, and smooth user
experience. In the end, choices and tradeoffs have to be made.

I would be delighted to spend more time on these questions and move forward
on nailing a more complete 'social networking over xmpp' spec, that would
evolve XEP-0277 to these use cases.


> 3) Can anyone try to 'follow' me and give me feedback ?
>> (xmpp:eschnou at vodafonernd.com?;node=urn%3Axmpp%3Amicroblog%3A0)
>>
>
> I have subscribed to your node and retrieved the items.
>
> Some <atom:entry/> contain two <atom:content/> elements, and I think it's
> not correct.
>
> Indeed. Will check that one.


> Also, the Pubsub items contain the ACL elements.
>
> Early implementation :-) By the way, it would also be nice to get this out
from the atom payload and a property of the publish operation. (as an
element in the <item/> or using publish-options).


> I'm to write an OSW component for ejabberd and have looked at the updated
> specs, but they seem to miss some informations on ACL. Is there any schema
> for ACL ?
>
>
Great ! Feel free to ping me directly if you want to discuss implementation.
Happy to have a phone call as well.. and we can even do it in french :-)

Thanks for the feedback. Lots of thing here, so maybe best to take these
various points one by one in individual posts and maybe discuss also offline
if you really want to move forward on these.

Laurent
JID: eschnou at vodafonernd.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.jabber.org/pipermail/standards/attachments/20100414/7549a260/attachment.html>


More information about the Standards mailing list