| You can look at this from two perspectives.
1) The marketing perspective
IP switching is available now. Tag switching is vapourware
2) The technical perspective.
IP switching (as currently defined and implemented) is confined to ATM networks.
Flows are identified dynamically by analysing the traffic flowing, and
when a "good" flow is detected, the VPI/VCI is set up to switch the flow. The
protocol used for this exchange is called IFMP. In its simplest forms each
separate flow requires a separate VPI/VCI. This scales poorly in many cases,
however there are variants whick allow some coarser grained flows and hence
reduce the number required. At the moment IP switching is for IP only,
but there is no real reason why it couldn't be made multiprotocol.
TAg switching in contrast covers all network technologies. It is not restricted
to ATM (and indeed scales poorly for ATM). The tag bindings are not set up
dynamically, but are pre-computed from the routing tables. Cisco claims that
this gives better scaling and isolation of function. The jury is still out.
In non-ATM technologies the tag is stuffed in a shim betweent he layer 2 and
layer 3 informatrion, and is re-written at each hop. i.e. you need new
hardware. For ATM they use VPI?VCI switching rather like IP switching, but
the tags are still set up in advance. Tag switching is designed to be
multiprotocol.
Really this whole subject is just too big to summarize in a few sentences.
Its not just IPsilon and Cisco playing here. There are all sorts of other
proposals. The big problem is that since very little (except some of the
Ipsilon stuff) is cast in solid product, you can pretty much say anything
is possible.
You need to do some reading. A good comparison can be found at
http://www.csl.sony.co.jp/person/demizu/inet/mlr.html
Mike
|