#2412 BLOC First D: Buddies not disappearing even when connection does

[Zarro Boogs per Child]

#2412: Buddies not disappearing even when connection does
-------------------------------+--------------------------------------------
  Reporter:  Zack              |       Owner:  smcv                  
      Type:  defect            |      Status:  new                   
  Priority:  blocker           |   Milestone:  First Deployment, V1.0
 Component:  presence-service  |     Version:                        
Resolution:                    |    Keywords:  collaboration         
  Verified:  0                 |  
-------------------------------+--------------------------------------------
Changes (by smcv):

 * cc: mors (removed)
 * cc: morgs, jg (added)


Comment:

 [Adding jg, kimquirk to Cc since we need an opinion on what the policy
 ought to be - please Cc UI people if needed]

 Gabble has a TCP connection to the server. When that TCP connection times
 out, it'll signal disconnection. There are really several things going on
 here:

 Issue 1: Does the Presence Service respond correctly to disconnections? -
 does it attempt to reconnect to Gabble? does it bring up Salut? does it
 signal that all buddies went offline? I'll have a look at the PS and see
 what it actually does, and whether we have a bug here.

 Issue 2: In an ideal world, how soon should a loss of connectivity cause a
 disconnection to be signalled by the CMs (and hence the above behaviour),
 i.e. how hard should we try to detect timeouts? This is a policy question
 on which we need input from UI/platform designers - jg's comment "I'd make
 the timeout just over an hour" conflicts rather with any attempt to detect
 disconnections immediately!

 Issue 3: How soon *does* a loss of connectivity cause a disconnection to
 be signalled by the CMs? Does it match (2)? If the conclusion of (2) is
 that we should try harder, possible things we could do include TCP
 keepalives as daf suggested, forcibly disconnecting Gabble in response to
 NetworkManager signals, etc.

-- 
Ticket URL: <https://dev.laptop.org/ticket/2412#comment:18>
One Laptop Per Child <https://dev.laptop.org>
OLPC bug tracking system