F5/unison
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update RELEASE_NOTES

Browse Source
This commit is contained in:
Tom Henderson
2016-03-12 17:37:01 -08:00
parent 6dbc04fe3d
commit 9941e75c5e
1 changed files with 91 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

121
RELEASE_NOTES
View File

@@ -10,6 +10,9 @@ Consult the file CHANGES.html for more detailed information about changed
API and behavior across ns-3 releases.
Release 3-dev
=============
Release 3.25
============
Availability
@@ -18,40 +21,92 @@ This release is not yet available.
Supported platforms
-------------------
To be determined.
This release has been tested on the following platforms:
- Ubuntu 15.10 (64 bit) with g++-5.2.1
- Ubuntu 14.04 (64 bit) with g++-4.8.2
- Fedora Core 23 (64 bit) with g++-5.3.1
- Fedora Core 22 (32 bit) with g++-5.1.1
- Fedora Core 21 (64 bit) with g++-4.9.2
- CentOS/RHEL 6.7 (64-bit) with g++-4.4.7
- OS X El Capitan 10.11.3 with Xcode 7.2.1 and Apple LLVM version 7.0.2
- OS X Yosemite 10.10.5 with Xcode 7.2.1 and Apple LLVM version 7.0.2
- FreeBSD 10.2-RELEASE (64 bit) with clang-3.4.1
New user-visible features
-------------------------
- (internet) The native TCP implementation has been refactored, for the following
improvements: 1) congestion control has been aligned with the Linux
architecture, to facilitate easier coding of new congestion control algorithms
or porting implementations from the Linux kernel, 2) Fast Retransmit and Fast
recovery algorithms, compliant to RFCs, are shared by all TCP variants, unlike
previous releases. 3) TCP Hybla and TCP HighSpeed have been added.
- (routing) The PrintRoutingTable* functions now print the following informations
each time they are called: 1) the node ID, 2) the global time, 3) the local time
4) the routing protocol type
- (internet) L4 protocols (e.g., TCP, UDP, ICMP, etc.) can be demultiplexed by IP
according to the incoming interface. In other words, it is now possible to
use specialized L4 protocols according to the interface.
- (internet) Ipv6Address::IsAllHostsMulticast() is now deprecated. FF02::3 is not
in the official reserved IPv6 multicast address list.
- (internet) Ipv6Address::IsAll[Nodes,Routers]Multicast() now checks the address
scope beyond the simple link-local. Nodes are checked for Interface-Local, Link-Local
and Realm-Local, Routers for the above plus Site-Local.
- (internet) Ipv6 routing protocols must now *not* forward packets to upper layers
unless for extremey specific cases. The Ipv6L3protocol handles almost all the
packets directed to the host.
- (internet) Ipv6 can now reject packets directed to an address not configured on
the interface they are received from (Strong End System Model, RFC 1222).
- (internet) The native TCP implementation has been refactored, for the
following improvements:
1) congestion control has been aligned with the Linux architecture,
to facilitate easier coding of new congestion control algorithms
or porting implementations from the Linux kernel,
2) Fast Retransmit and Fast Recovery algorithms, compliant to RFCs, are
shared by all TCP variants, unlike previous releases.
3) TCP Hybla and TCP HighSpeed variants have been added.
- (routing) The PrintRoutingTable* functions now print the following information
each time they are called:
1) the node ID,
2) the global time,
3) the local time (allowing for future introduction of per-node clocks),
4) the routing protocol type
- (traffic control) A traffic control sublayer has been added with the aim of
introducing an equivalent of the Linux Traffic +Control infrastructure into
ns-3. This layer sits in between the NetDevices (L2) and the network
protocol (e.g. IP). It is in charge of processing packets and performing
actions on them: scheduling, dropping, marking, policing, etc. Initial
support for flow control has been added to the PointToPointNetDevice;
backward-compatible support for other device types is provided.
- (traffic control) A three band priority queue modelled after Linux
pfifo_fast queueing discipline has been added.
- (traffic control) Adaptive RED extensions have been added to the RED model
- (wifi) A number of changes related to improving support for 802.11n/ac
models have been introduced.
1) Two rate controls compatible with High Throughput (HT) Wi-Fi modes
(e.g. 802.11n/ac), Ideal and MinstrelHT, have been added
2) Support for multiple spatial streams (MIMO) using existing SISO
error models
3) The WiFi helper API for the WifiMac sublayer, and the configuration
of HT modes and parameters, has been simplified
4) Better support for backward compatibilty with legacy modes
- (wifi) 802.11g was made backward compatible with 802.11b networks
- (internet) L4 protocols (e.g., TCP, UDP, ICMP, etc.) can be demultiplexed
by IP according to the incoming interface. In other words, it is now
possible to use specialized L4 protocols according to the interface.
- (internet) Ipv6Address::IsAllHostsMulticast() is now deprecated.
FF02::3 is not in the official reserved IPv6 multicast address list.
- (internet) Ipv6Address::IsAll[Nodes,Routers]Multicast() now checks the
address scope beyond the simple link-local. Nodes are checked for
Interface-Local, Link-Local and Realm-Local, Routers for the above plus
Site-Local.
- (internet) Ipv6 routing protocols must now *not* forward packets to upper
layers unless for extremey specific cases. The Ipv6L3protocol handles
almost all the packets directed to the host.
- (internet) Ipv6 can now reject packets directed to an address not configured
on the interface they are received from (Strong End System Model, RFC 1222).
- (internet) UDP and Ipv6 RAW sockets can now join Ipv6 multicast groups.
Incomin packets will be filtered according to if there's a socket listening to
that group. Source filtering is left to the application.
- (network) ns-3 is now capable of serializing cooked (SLL) headers. This is used
in DCE to allow the generation of pcap directly readable by wireshark.
- (internet) It is now possible to set custom values for RipNg Link Down (standard is 16).
Incoming packets will be filtered according to if there's a socket listening
to that group. Source filtering is left to the application.
- (network) ns-3 is now capable of serializing cooked (SLL) headers. This is
used in DCE to allow the generation of pcap directly readable by wireshark.
- (internet) It is now possible to set custom values for RipNg Link Down
(standard is 16).
- (internet) permanent (static) NDISC entries can be created for IPv6
- (internet) IPv4 RIPv2 protocol is now available.
- (internet) IPv4 RIPv2 protocol model is now available.
Bugs fixed
----------
@@ -65,6 +120,7 @@ Bugs fixed
- Bug 2003 - Missing DSSS short PLCP preamble
- Bug 2041 - TCP RTO needs unit tests
- Bug 2068 - Timestamp option conforms to RFC 7323
- Bug 2102 - Make global routing robust to bridged links
- Bug 2116 - refactoring aggregation API
- Bug 2120 - 802.11g networks are not compatible with 802.11b clients
- Bug 2141 - TCP DataSent callback now correctly notifies sent data, without missing bytes in particular conditions
@@ -72,13 +128,15 @@ Bugs fixed
- Bug 2159 - TCP advertises wrong receive window
- Bug 2176 - Building IPv4 address from char* doesn't look reliable
- Bug 2183 - LiIonEnergySourceHelper is not in the energy wscript
- Bug 2184 - Integer overflow in Wi-Fi MacLow; remove dependency between tx and rx reference number field in MPDU status
- Bug 2185 - WiFi MacLow may respond to errored frames that it should ignore
- Bug 2193 - Simulation fails when transmitting very small MPDU subframes
- Bug 2195 - Udp[*]Client can't send packets to broadcast address
- Bug 2201 - Simulation fails when active probing is used in 802.11n/ac with channel bonding enabled
- Bug 2207 - Print node ID and time when printing routing tables
- Bug 2206 - Split internet-apps from applications
- Bug 2207 - Print node ID and time when printing routing tables
- Bug 2208 - Interface index based L4 protocols
- Bug 2210 - set Order field to 0 for HT/VHT
- Bug 2211 - Ipv{4,6}EndPoint can cause memory corruption
- Bug 2213 - Inconsistencies may exist between the selected WifiMacHelper and the chosen 802.11 version
- Bug 2219 - SixLowPanNetDevice hangs trying to decode a IPv6 Fragment extension header
@@ -90,6 +148,7 @@ Bugs fixed
- Bug 2243 - TCP Socket Fork() fails to copy some parameters, causing connections to close prematurely on retransmit
- Bug 2246 - Some DSR LogComponents and classes are not defined in a unique way.
- Bug 2247 - Disabled Fast retransmit after an RTO
- Bug 2249 - LTE perl speed tests possibly broken
- Bug 2252 - Nix needs StaticRouting to work
- Bug 2254 - Ipv[4,6]RawSocket can return the wrong number of bytes sent.
- Bug 2255 - Ipv6RawSocket does not call data sent callbacks.
@@ -109,6 +168,7 @@ Bugs fixed
- Bug 2287 - FlowMonitor is confused by IP over IP.
- Bug 2288 - Ipv4 broadcast and multicast packets are replicated on all the interfaces.
- Bug 2289 - RIP v2 implementation
- Bug 2290 - Cannot run click examples from test.py
- Bug 2292 - Uninitialized variables since commit 7c60a9f8f271
- Bug 2293 - Red Queue Estimator spins when trying to compute queue average size under long idle times.
- Bug 2297 - 802.11n/ac networks are not compatible with legacy clients
@@ -127,6 +187,7 @@ Bugs fixed
- Bug 2321 - Wifi rate managers should not be triggered for each MPDU when A-MPDU is used
- Bug 2327 - CWmin value selection for 802.11g is not compliant with the standard rules
- Bug 2328 - changes to antenna configuration should update HT MCS set
- Bug 2331 - Wrong mobility model in vanet-routing-compare example
- Bug 2332 - Assert failed when using aggregation and two pairs of AP-STA
Known issues