Differences between current version and predecessor to the previous major change of CSMA/CA.

Other diffs: Previous Revision, Previous Author, or view the Annotated Edit History

Newer page: version 11 Last edited on Wednesday, July 20, 2005 5:02:07 pm by MattBrown
Older page: version 9 Last edited on Monday, November 22, 2004 2:22:41 pm by MikeBeattie Revert
@@ -3,16 +3,8 @@
 A network access protocol used in Apple's LocalTalk networks and wireless network access methods. It is a variation of the [CSMA/CD] protocol used in [Ethernet] networks. As in [CSMA/CD] method, each device listens to the signal level to determine when the channel is idle. Unlike the [CSMA/CD], it, then, waits for a random amount of time before trying to send a packet. After a while, the device senses the signal level again and if the channel is free, the packet is sent. If the channel is busy, the time interval before the next attempt is doubled. In some versions of the protocol the time interval between attempts is based on the device's position in the devices list. 
 The principal medium access method employed by IEEE 802.11 wireless [LAN]s ([WLAN]s). It is a ""listen before talk"" method of minimizing (but not eliminating) collisions caused by simultaneous transmission by multiple radios. IEEE 802.11 states the collision avoidance method rather than the collision detection that must be used, because the standard employs half-duplex radios-radios capable of transmission or reception-but not both simultaneously. 
-__NOTE__: Although the above description may be true, the more commonly accepted description of [CSMA/CA] is attached to TokenRing networks. Token ring achieves collision avoidance by only allowing a computer which is currently in possession of the token to transmit. In such a configuration, there is never the possiblity of a collision. This explains why in many cases 16MBit token ring ( [CSMA/CA]) can easily outperform 100MBit [Ethernet] ([CSMA/CD] ).  
-Also, contrary to popular misconception, it is possible ( and common) to use more than one token on the network. The data traversing the network still avoids collisions by only allowing transfer of data in one direction around the ring. While this may sound inefficient, it is in practice quite an impressive performer.  
-What is your source for the above information? Clearly it depends on network load factor. 100MBps Ethernet can transfer data at a blistering rate compared to 16MBps TokenRing, and with substantially less latency to boot. Coupled with a multi-port ''EthernetSwitch'' (and not an EthernetHub), total network aggregate bandwidth can achieve many times this rate, perhaps even going into the gigabits per second, for switches with suitably numerous ports.  
-Also, TokenRing is __not__ CSMA/CA. There is no carrier sensing at all -- if you don't have the token, you have no need to sense a carrier, because you cannot transmit (legally) on the network at all. TokenRing is a variation of DAMA (DemandAssignedMultipleAccess), which can also be applied to bus-topologies as well as star-topologies (e.g., DAMA has proven itself quite well in an amateur packet radio environment, even despite the 1200bps and 9600bps link rates)
+You may also like to compare [CSMA/CA] with DAMA (DemandAssignedMultipleAccess ) which is used for protocols such as TokenRing and works in a similar fashion to [ CSMA/CA]