Cell Relay Archive

[Saso STOJANOVSKI <sassos@res.enst.fr>]

alan,


> I think you may be confusing ATM with frame relay.  There is an end-to-end
> reliable transport protocol defined for ATM called SSCOP (Service Specific
> Connection Oriented Protocol) that is an SSCS implementation (It is
> specified in the Signalling AAL or SAAL) for the ATM signaling application
> along with AAL5.  However, there is still no such mechanism defined for
> frame relay.  So while the ATM core can have congestion/flow control and
> end-to-end recovery, the frame relay access cannot.


i think you are wrong.

the q.922 recommendation specifies two sublayers: lapf core and lapf
control.

whereas lapf core may be compared to the atm layer in atm networks, lapf
control operates on an end-to-end basis and is directly comparable to
sscop.

have a look at the protocol stacks:


      lapf control                                    lapf control
      lapf core         lapf core     lapf core       lapf core
      phy               phy           phy             phy


      aal (sscop)                                     aal (sscop)
      atm               atm           atm             atm
      phy               phy           phy             phy



lapf control is basically the same sliding window protocol as in lapb.

if there were applications directly accessing the lapf control, one could
consider it as a transport protocol since it operates on end-to-end basis.

as far as i know, there are no applications directly accessing the
frame relay service, since the "end user" from the viewpoint of the frame
relay networks is the frad.

so, since frame relay networks carry mostly lan traffic and ip traffic,
nobody uses the lapf control in the user plane. tcp will anyway take
care of assuring the transport.

however, lapf control is mandatory in the frame relay *control* plane.

but then again, since there are no switched frame relay networks out
there, nobody cares about lapf control.

note that the same thing applies to atm. there are no native data
applications for atm, and nobody uses sscop in the user plane. but sscop
is used in the control plane.




i would like to know whether lapf control is used in the frame relay's 
user plane when carrying x.25 traffic over frame relay? 

x.25 assumes that the data-link connection (in this case the frad-to-frad
connection) is reliable, doesn't it? can somebody answer this point?


--sS