powered by Jive Software

Share Roster user informed of roster deletion over and over again

I was hoping that someone might be able to point me in the right direction to solve a problem or at least give me the current state of a solution. I have searched the forms and found some threads but none of them give any type of answer.

Here is the problem: I am using Wildfire 2.4.4 with Soapbox clients (Pandion derivative) and LDAP (Active Directory) users and authentication. If I remove someone from a shared group it is broadcast to all users that the user has removed the person from his contact list. The problem is that some users see this every time they start the client back up. Its not all the users but so far at least 5 out of 72 have reported it to me. It does not happen me but it also does not happen to other users that are not admins.

There doesn’'t seem to be any simularity between the 5 people/computers that have informed me of this problem. They are from different groups (groups that are not shared on the roster and are used just for the broadcast plugin). Some of the people only have a few contacts that are informed have removed them from their contact list while others have almost all their contacts this way.

We use one main group called “Everybody” that is shared and has all the users in it. The other shared group is called “Broadcast” and has all the broadcast groups in it i.e. all@broadcast.example.com.

Thank you for any help you can give me in this problem.

Addtional Information:

The problem only occurs if the account uses a Pandion client (2.1 to 2.5 tried).

The problem is not a computer issue it will happen on a computer that works fine with one account but with another account (Wildfire account) the client will display the problem.

Message was edited by: mhindt

Message was edited by: mhindt

Hi Mark,

there was (is ?) an issue with “special” characters like “^” in the username/resource, so Wildfire could not delete the entry. You are using LDAP, I’‘m not so I don’'t know if Wildfire stores something also to the local database (eg to notify the users).

Wildfire could also have a caching problem with LDAP, so a restart of Wildfire could solve this problem.

Do you have a tool (toad, phpmyadmin, oem, dbvisualizer) to view the database tables? Maybe you can identify a similar problem in the JIVEOFFLINE table.


Hi LG,

Thanks for the reply. Restarting of the Wildfire did not fix the problem. I looked at the database and don’'t see anything in the JIVEOFFLINE table for any of the people that have this issue. I could disable Offline message store its pretty empty.

One thing I did notice is that if the user who have the problem re-request authorization the problem goes away for that contact but as it is 70+ contacts for each person it is a bit anoying. I have for the mean time changed clients for the affected users from the Pandion based client to Exodus. I do not want to have more then one offical IM client for the company and Exodus’'s lack of tabbed chat windows and generally more choices in setup leave it less desirable of a client.

I’'m going to post another thread asking if it is possible to migrate from the embeded db to MS SQL or some other RDBS with out having to re-enter all the shared groups. If I can look at the database live my be I can find where the problem is.

Hey Mark,

When removing a user from a shared group or making some modification to the group configuration that implies removing existing subscrptions the server will send unsubscription notifications to the affected user. Usually clients display a dialog to the user informing that his subscription to the given contact was removed and once the dialog is closed (or was opened) the client will send a notification to the server informing that the user was notified of such event. However, if the client never sends such notification back to the server then the server will continue sending the unsubscription notification to the client until the client acks.

This functionality is better explained in http://www.xmpp.org/specs/rfc3921.html#int-cancel step 4. You may want to check if the client that you were using is in fact sending the ack to the server. Let me know if the server is still sending the notification after the ack.


– Gato

Hi Gaston,

First let me thank you and the rest of the Wildfire team for making such a great server.

The unsubsription events do not happen with the other 60 some users besides the intial unsubscription notifcations. Also for the computers that do properly send ack (I’‘m assuming) with the Pandion Client, mine for instants, if I logon to wildfire using the affected user’‘s credential it displays the same problem of always notifing the client that the contact has remove the user from the contact’'s contact list.

Let me run you through what is happening

User X has the issue of the unsubscriptions events always being fire when he starts his client.

User Y does not suffer this problem though he uses the same client.

If User X logs in using user Y’'s computer the problem is still there even through multiple logins.

If User Y uses user X’'s computer the problem still does not appear.

If User X uses a different client besides Pandion the problem does not appear but if the user goes back to Pandion it reappears. I have captured the packets from the end users machine using Ethereal. I can email them to you if that would help but I cannot post them because they are a bit sensitive in nature.

Hey Mark,

You can send me the packets so I can review them. BTW, try to include the original packet sent from the server to the client when the group was modified and further exchanged presences between the client and the server.


– Gato

I finnally used Spark to logon to one of the effected accounts then tried Pandion again and the issue was not present.