Even more recent
----------------
if NOTIFY doesn't show up, set some flag to force reSUB next time around.
if NOTIFY doesn't show, print out more info about the SUB. Currently
   cannot get to the SubSession to show/manip

- For stale NOTIFY, display last error code
- Remove SUB from DB upon response error
- Look for empty NOTIFY and/or status line and remove from DB
- Look for error codes to PUB response and remove from DB
- Why is expire time 0 for SUBs?

Latest ToDo for RENDv2/Pres1Sketch
---------------------
- activate all repeats of PUB at same time. SUB code assumes this; need
  to make true
- Add mechanism to fail late notifies based upon timeout (or user input?)
- For pending count delays, add in expect NOTIFYs count, not just PUB count
- For throttling on maxwork, use expected NOTIFY count
- Add mechanism to force adjusting of PUBS and/or SUBs even when no
  level change (cycle thru accounts)
- make the pending limit variable (treat like level/maxRate)
- fix shutdown (clean out the PendReq/PendNot queue)
- add more aggressive shutdown modes (wait less)
- better handling of failed requests

ToDo-REND
---------
1. Switch to pure outbound mode
2. Fix 'force-transport' handling
4. Why do we get in repro
	In RequestContext, after processing a sip response: all transactions are terminated, but we have not sent a final response. (What happened here?)".
    Probably related to unreachable flow.
4b. Make test program to generate.
5. Need periodic garbage collection dead transports.
6. Ask review of epoll code and merge
7. Address lost connections, keep-alive handling, re-trigger new dialogs
   when lost? Try out handling LostConnection message

8. Make ConnectionManager.mHead a pointer so that Connection.hxx isn't 
   included, so that FdPoll.hxx isn't forced in at top level

9. In Dlg, make sure don't try to PUB e=0 without etag. (assert)

Add logic to change called (re-start dialogs)

add REGISTER support

get reNEW mode working

add infinite loop mode with console mode commands (sort of like sipp)

detect conn fail and restart dialogs. Need mapping in ka to individual dlgs.

dither timing periods (reNEW & ka?)

eliminate as many calls to random() as possible. Use counting serials for callid, etc.

check for excess copies of sip messages and eliminate

outbound mode
   * send to specific flow when in-dlg


Todo-resip/stack
----------------
9. Make sure UDP (esp) reads all available packets in loop.
   May need to add some limiting on queue size. If so, may need
   to implement some sort of "pause" method to pause FileIO for reading,
   and then to wake everyone up when unpaused.

Add SipStackOptions class and integrate, especially with PollThread

Rename whole "Poll" concept to something else. Eventloop, or just event.


