Changes in 1.3.2 (May 2006)
---------------------------
* Fixed bug
* Added -t option to kill pathload_rcv after user specified time

Changes in 1.3.1 (May 2005)
---------------------------
* Modified to work in 64bit environment
  - Thanks to Omid Khalili and Shava Smallen for testing and reporting several
  bugs in 64bit environment
* Modified to work without thread library
* Bug fixes
  - for-loop in adjust_offset_to_zero( ) should go upto len-1
  
Changes in 1.3.0 (Sep 2004)
---------------------------
* Initial probing rate, when ADR estimate is 0, is changed from min_rate to 
  avg of min_rate and max_rate. 
* Bug fixes
  - Moved send_latency call after pathload_snd determines max_pkt_sz.
  This bug resulted in packet size of 0 being used in send_latency function.
  - min_timer_intr set to zero in the pathload_snd_fns.c::send_fleet() code.  
  This caused floating point exceptions (when it was zero).
  - Changed max_pkt_sz to cur_pkt_sz in recv_fleet function. This caused 
  segmentation fault in certain cases (when trying to free memory).
  - Modified pathload sender code such that sender exits (or enters wait state
  for new receiver) if it receives either cur_pkt_sz, time_interval or probe_rate
  equal to 0.
  

Changes in 1.2.0 (Apr 2004)
---------------------------
* Interrupt coalescence changes
  Detects IC and uses alternate algorithm (Published in PAM '04).

Changes in 1.1.1 (Jan 2004)
---------------------------
* Bug fixes.
  - Make new thread detached in pathload receiver. This fixes a bug in older
    version, where receiver would hang after approx 20 fleet as not enough
    memory was available to create a new thread and receiver would exit with a
    message "Receiver not responding".
  - Fixed a bug in sender that would cause sender running in iterative mode to
    exit with "Broken pipe" if receiver was killed.
  - Simplified code to detect interrupt coalescence.
  - Several other minor changes in the code

Changes in 1.1.0 (Apr 2003)
---------------------------
* Added option to have output in netlogger format.
* Implemented function to measure minimum sleep time using select.
* Added select so that sender doesnot block and now waits a maximum of 50 sec
  for a control message from receiver during a ongoing measurement.  
* Reduced the aggregate threshold from .7 -> to .6.
* Receiver uses different thread for control messages while getting probing
  packets in main thread. Don't use select anymore.
