CHANGES.html
author Mathieu Lacage <mathieu.lacage@sophia.inria.fr>
Wed, 10 Dec 2008 01:34:04 -0800
changeset 4002 a12900ea255e
parent 3982 4f370a1b637c
child 4260 29587ce02a57
permissions -rw-r--r--
rescan python

<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
   <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
   <title>ns-3 Change Log</title>
</head>
<body>

<h1>
ns-3: API and model change history</h1>
<!--
This ChangeLog is updated in the reverse order 
with the most recent changes coming first.  Date format:  DD-MM-YYYY
-->
<p>
ns-3 is an evolving system and there will be API or behavioral changes
from time to time.   Users who try to use scripts or models across
versions of ns-3 may encounter problems at compile time, run time, or
may see the simulation output change.  </p>
<p>
We have adopted the development policy that we are going to try to ease
the impact of these changes on users by documenting these changes in a
single place (this file), and not by providing a temporary or permanent
backward-compatibility software layer.  </p>
<p>
The goal is that users who encounter a problem when trying to use older
code with newer code should be able to consult this file to find
guidance as to how to fix the problem.  For instance, if a method name
or signature has changed, it should be stated what the new replacement
name is. </p>
<p>
Note that users who upgrade the simulator across versions, or who work
directly out of the development tree, may find that simulation output
changes even when the compilation doesn't break, such as when a
simulator default value is changed.  Therefore, it is good practice for
_anyone_ using code across multiple ns-3 releases to consult this file,
as well as the RELEASE_NOTES, to understand what has changed over time.
</p>
<p>
This file is a best-effort approach to solving this issue; we will do
our best but can guarantee that there will be things that fall through
the cracks, unfortunately.  If you, as a user, can suggest improvements
to this file based on your experience, please contribute a patch or drop
us a note on ns-developers mailing list.  </p>

<hr>
<h1>changes from ns-3.2 to ns-3.3</h1>

<h2>new API:</h2>
<ul>
<li> ns-3 ABORT macros in src/core/abort.h</li>
<li> Config::MatchContainer </li>
<li> ConstCast and DynamicCast helper functions for Ptr casting</li>
<li> StarTopology added to several topology helpers </li>
<li> NetDevice::IsBridge () </li>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/4c1c3f6bcd03">4c1c3f6bcd03</a></li>
<ul>
<li>
The PppHeader previously defined in the point-to-point-net-device code has been 
made public.
</li>
</ul>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/16c2970a0344">16c2970a0344</a></li>
<ul>
<li>
An emulated net device has been added as enabling technology for ns-3 emulation
scenarios.  See src/devices/emu and examples/emu-udp-echo.cc for details.
</li>
</ul>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/4222173d1e6d">4222173d1e6d</a></li>
<ul>
<li>
Added method InternetStackHelper::EnableAsciiChange to allow allow a user to 
hook ascii trace to the drop trace events in Ipv4L3Protocol and ArpL3Protocol.
</li>
</ul>

</ul>
<h2>changes to existing API:</h2>
<ul>

<li> NetDevice::MakeMulticastAddress() was renamed to NetDevice::GetMulticast()
and the original GetMulticast() removed </li>

<li> Socket API changes:
<ul>
<li> return type of SetDataSentCallback () changed from bool to void </li>
<li> Socket::Listen() no longer takes a queueLimit argument</li>
</ul>

<li> As part of the Wifi Phy rework, there have been several API changes
at the low level and helper API level.  </li>
<ul>
<li>  At the helper API level, the WifiHelper was split to three classes: 
a WifiHelper, a YansWifiChannel helper, and a YansWifiPhy helper.  Some
functions like Ascii and Pcap tracing functions were moved from class
WifiHelper to class YansWifiPhyHelper. 
<li>  At the low-level API, there have been a number of changes to
make the Phy more modular:</li>
<ul>
<li> composite-propagation-loss-model.h is removed</li>
<li> DcfManager::NotifyCcaBusyStartNow() has changed name</li>
<li> fragmentation related functions (e.g. DcaTxop::GetNFragments()) have
changed API to account for some implementation changes</li>
<li> Interference helper and error rate model added </li>
<li> JakesPropagationLossModel::GetLoss() moved to PropagationLoss() class</li>
<li> base class WifiChannel made abstract </li>
<li> WifiNetDevice::SetChannel() removed </li>
<li> a WifiPhyState helper class added </li>
<li> addition of the YansWifiChannel and YansWifiPhy classes </li>
</ul>
</ul>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/dacfd1f07538">dacfd1f07538</a></li>
<ul>
<li>
Change attribute "RxErrorModel" to "ReceiveErrorModel" in CsmaNetDevice for 
consistency between devices.
</li>
</ul>

</ul>
<h2>changed behavior:</h2>
<ul>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/ed0dfce40459">ed0dfce40459</a></li>
<ul>
<li>
Relax reasonableness testing in Ipv4AddressHelper::SetBase to allow the 
assignment of /32 addresses.
</li>
</ul>

<li>17-11-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/756887a9bbea">756887a9bbea</a></li>
<ul>
<li>
Global routing supports bridge devices.
</li>
</ul>
</ul>

<hr>
<h1>changes from ns-3.1 to ns-3.2</h1>

<h2>new API:</h2>
<ul>

<li>26-08-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/5aa65b1ea001">5aa65b1ea001</a></li>
<ul>
<li>
Add multithreaded and real-time simulator implementation.  Allows for emulated
net devices running in threads other than the main simulation thread to schedule
events.  Allows for pacing the simulation clock at 1x real-time.
</li>
</ul>


<li>26-08-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/c69779f5e51e">c69779f5e51e</a></li>
<ul>
<li>
Add threading and synchronization primitives.  Enabling technology for 
multithreaded simulator implementation.
</li>
</ul>

</ul>
<h2>new API in existing classes:</h2>
<ul>

<li>01-08-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/a18520551cdf">a18520551cdf</a></li>
<ul>
<li>class ArpCache has two new attributes:  MaxRetries 
and a Drop trace.  It also has some new public methods but these are 
mostly for internal use.
</ul>
</li>

</ul>
<h2>changes to existing API:</h2>
<ul>

<li>05-09-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/aa1fb0f43571">aa1fb0f43571</a></li>
<ul>
<li>
Change naming of MTU and packet size attributes in CSMA and Point-to-Point devices<br>
After much discussion it was decided that the preferred way to think about 
the different senses of transmission units and encapsulations was to call the 
MAC MTU simply MTU and to use the overall packet size as the PHY-level attribute
of interest.  See the Doxygen of CsmaNetDevice::SetFrameSize and 
PointToPointNetDevice::SetFrameSize for a detailed description.
</li>
</ul>

<li>25-08-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/e5ab96db540e">e5ab96db540e</a></li>
<ul>
<li>
bug 273: constify packet pointers.<br>
The normal and the promiscuous receive callbacks of the NetDevice API
have been changed from:
<pre>
Callback<bool,Ptr<NetDevice>,Ptr<Packet>,uint16_t,const Address &>
Callback<bool,Ptr<NetDevice>, Ptr<Packet>, uint16_t,
         const Address &, const Address &, enum PacketType >
</pre>
to:
<pre>
Callback<bool,Ptr<NetDevice>,Ptr<const Packet>,uint16_t,const Address &>
Callback<bool,Ptr<NetDevice>, Ptr<const Packet>, uint16_t,
         const Address &, const Address &, enum PacketType >
</pre>
to avoid the kind of bugs reported in 
<a href="http://www.nsnam.org/bugzilla/show_bug.cgi?id=273">bug 273</a>.
Users who implement a subclass of the NetDevice base class need to change the signature
of their SetReceiveCallback and SetPromiscReceiveCallback methods.
</li>
</ul>


<li>04-08-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/cba7b2b80fe8">cba7b2b80fe8</a></li>
<ul>
<li>
Cleanup of MTU confusion and initialization in CsmaNetDevice<br>
The MTU of the CsmaNetDevice defaulted to 65535.  This did not correspond with
the expected MTU found in Ethernet-like devices.  Also there was not clear 
documentation regarding which MTU was being set.  There are two MTU here, one
at the MAC level and one at the PHY level.  We split out the MTU setting to make
this more clear and set the default PHY level MTU to 1500 to be more like
Ethernet.  The encapsulation mode defaults to LLC/SNAP which then puts the
MAC level MTU at 1492 by default.  We allow users to now set the encapsulation
mode, MAC MTU and PHY MTU while keeping the three values consistent.  See the
Doxygen of CsmaNetDevice::SetMaxPayloadLength for a detailed description of the
issues and solution.
</li>
</ul>

<li>21-07-2008; changeset 
<a href="
http://code.nsnam.org/ns-3-dev/rev/99698bc858e8">99698bc858e8</a></li>
<ul>
<li> class NetDevice has added a pure virtual method that
must be implemented by all subclasses:
<pre>
virtual void SetPromiscReceiveCallback (PromiscReceiveCallback cb) = 0;
</pre>
All NetDevices must support this method, and must call this callback
when processing packets in the receive direction (the appropriate place
to call this is device-dependent).  An approach to stub this out
temporarily, if you do not care about immediately enabling this
functionality, would be to add this to your device:
<pre>
void
ExampleNetDevice::SetPromiscReceiveCallback
(NetDevice::PromiscReceiveCallback cb)
{ 
  NS_ASSERT_MSG (false, "No implementation yet for
SetPromiscReceiveCallback");
}
</pre>
To implement this properly, consult the CsmaNetDevice for examples of
when the m_promiscRxCallback is called.
</li>
</ul>

<li>03-07-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/d5f8e5fae1c6">d5f8e5fae1c6</a></li>
<ul>
<li>
Miscellaneous cleanup of Udp Helper API, to fix bug 234
<pre>
class UdpEchoServerHelper
{
public:
- UdpEchoServerHelper ();
- void SetPort (uint16_t port); 
+ UdpEchoServerHelper (uint16_t port);
+ 
+ void SetAttribute (std::string name, const AttributeValue &value);
ApplicationContainer Install (NodeContainer c);

class UdpEchoClientHelper
{
public:
- UdpEchoClientHelper ();
+ UdpEchoClientHelper (Ipv4Address ip, uint16_t port);
- void SetRemote (Ipv4Address ip, uint16_t port);
- void SetAppAttribute (std::string name, const AttributeValue &value);
+ void SetAttribute (std::string name, const AttributeValue &value);
ApplicationContainer Install (NodeContainer c);
</pre>
</li>
</ul>

<li>03-07-2008; changeset 
<a href="
http://code.nsnam.org/ns-3-dev/rev/3cdd9d60f7c7">3cdd9d60f7c7</a></li>
<ul>
<li>
Rename all instances method names using "Set..Parameter" to "Set..Attribute"
(bug 232)
</li>
<li> How to fix your code:  Any use of helper API that was using a method
"Set...Parameter()" should be changed to read "Set...Attribute()".  e.g.
<pre>
- csma.SetChannelParameter ("DataRate", DataRateValue (5000000));
- csma.SetChannelParameter ("Delay", TimeValue (MilliSeconds (2)));
+ csma.SetChannelAttribute ("DataRate", DataRateValue (5000000));
+ csma.SetChannelAttribute ("Delay", TimeValue (MilliSeconds (2)));
</pre>
</li>
</ul>
</li>

</ul>
<h2>changed behavior:</h2>
<ul>

<li>07-09-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/5d836ab1523b">5d836ab1523b</a></li>
<ul>

<li>
Implement a finite receive buffer for TCP<br>
The native TCP model in TcpSocketImpl did not support a finite receive buffer.
This changeset adds the following functionality in this regard:
<ul>
<li>
Being able to set the receiver buffer size through the attributes system.
</li>
<li>
This receiver buffer size is now correctly exported in the TCP header as the
advertised window.  Prior to this changeset, the TCP header advertised window
was set to the maximum size of 2^16 bytes.
window
</li>
<li>
The aforementioned window size is correctly used for flow control, i.e. the
sending TCP will not send more data than available space in the receiver's
buffer.
</li>
<li>
In the case of a receiver window collapse, when a advertised zero-window
packet is received, the sender enters the persist probing state in which
it sends probe packets with one payload byte at exponentially backed-off
intervals up to 60s.  The reciever will continue to send advertised 
zero-window ACKs of the old data so long as the receiver buffer remains full.
When the receiver window clears up due to an application read, the TCP
will finally ACK the probe byte, and update its advertised window appropriately.
</li>
</ul>
See 
<a href="http://www.nsnam.org/bugzilla/show_bug.cgi?id=239"> bug 239 </a> for
more.
</li>
</ul>

<li>07-09-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/7afa66c2b291">7afa66c2b291</a></li>
<ul>
<li>
Add correct FIN exchange behavior during TCP closedown<br>
The behavior of the native TcpSocketImpl TCP model was such that the final
FIN exchange was not correct, i.e. calling Socket::Close didn't send a FIN
packet, and even if it had, the ACK never came back, and even if it had, the
ACK would have incorrect sequence number.  All these various problems have been
addressed by this changeset.  See 
<a href="http://www.nsnam.org/bugzilla/show_bug.cgi?id=242"> bug 242 </a> for
more.
</li>
</ul>

<li> 28-07-2008; changeset 
<a href="http://code.nsnam.org/ns-3-dev/rev/6f68f1044df1">6f68f1044df1</a>
<ul>
<li>
OLSR: HELLO messages hold time changed to 3*hello
interval from hello interval.  This is an important bug fix as
hold time == refresh time was never intentional, as it leads to
instability in neighbor detection.
</ul>
</li>

</ul>

</body>
</html>