RE: [RPRWG] SendA, SendB, SendC
John,
Consider the following scenario. A station has
enough Class A credits (i.e. SendA is asserted).
The client makes a request for transmission.
However, there's a train of transit packets on
the transit path because of which the packet
from the client cannot be put on the ring right
away. So, where is the packet sitting while
waiting for insertion - in the MAC or in the
client?
If, during this time, the packet is sitting in the
client, it could just as well have been the case
that the SendA signal was unasserted and the
the MAC was building up its credits for Class A.
So I don't see how having the signals would help
because it can't avoid the above scenario.
If the packet sits in the MAC while waiting for
transmission, then there is some value in providing
these signals. I was just wondering what other
MACs do.
-Anoop
> -----Original Message-----
> From: John Lemon [mailto:JLemon@xxxxxxxxxxxx]
> Sent: Wednesday, June 19, 2002 2:20 PM
> To: 'Anoop Ghanwani'; 'stds-802-17@xxxxxxxx'
> Subject: RE: [RPRWG] SendA, SendB, SendC
>
>
> Anoop,
>
> It is behaviorally different in that a (semi-) intelligent
> client can use
> these signals to plan what to send, rather than send and hope
> it works.
> Also, since the physical interface from the client to the MAC is not
> specified, there is no guarantee that the client even knows
> when an attempt
> to send fails.
>
> jl
>
> -----Original Message-----
> From: Anoop Ghanwani [mailto:anoop@xxxxxxxxxxxxxx]
> Sent: Wednesday, June 19, 2002 2:04 PM
> To: 'John Lemon'; Anoop Ghanwani; 'stds-802-17@xxxxxxxx'
> Subject: RE: [RPRWG] SendA, SendB, SendC
>
>
>
> John,
>
> That still doesn't answer my question of how this is
> behaviorally different from the MAC client choosing
> to heed these signals. Regardless of whether it heeds
> the signals or not, the traffic will be blocked. In
> that case, I don't see the need for the signals. If
> it enables some kind of functionality, then I'd be all
> all for it. Otherwise, were asking for something in
> the MAC that's essentially redundant.
>
> -Anoop
>
> > -----Original Message-----
> > From: John Lemon [mailto:JLemon@xxxxxxxxxxxx]
> > Sent: Wednesday, June 19, 2002 2:09 PM
> > To: 'Anoop Ghanwani'; 'stds-802-17@xxxxxxxx'
> > Subject: RE: [RPRWG] SendA, SendB, SendC
> >
> >
> > Anoop,
> >
> > I'm not enough of a token ring expert to answer your
> > questions about how and
> > why 802.5 works the way it does. Maybe our esteemed vice
> > chair can answer
> > your question about how and why 802.5's treatment of
> > priorities is different
> > from 802.17's usage of service classes.
> >
> > As for what happens if a client chooses to ignore the
> > signals, the client's
> > frames may not be accepted by the MAC. The send signals are a
> > friendly way
> > of informing the client that it has exceeded the relevant
> > shaper. If it
> > tries to send a packet with a Service Class value of X
> while the sendX
> > signal is not asserted, the MA_UNITDATA.request will be refused.
> >
> > jl
> >
> > -----Original Message-----
> > From: Anoop Ghanwani [mailto:anoop@xxxxxxxxxxxxxx]
> > Sent: Wednesday, June 19, 2002 1:30 PM
> > To: 'John Lemon'; Anoop Ghanwani; 'stds-802-17@xxxxxxxx'
> > Subject: RE: [RPRWG] SendA, SendB, SendC
> >
> >
> >
> > John,
> >
> > 802.5 MACs have several priorities and the client
> > can request a specific priority for its transmission.
> > Why is 802.17 any different?
> >
> > If the client can ignore the signals, why do we
> > need to have them? In other words, what happens
> > different if the client ignores them versus
> > uses them.
> >
> > -Anoop
> >
> > > -----Original Message-----
> > > From: John Lemon [mailto:JLemon@xxxxxxxxxxxx]
> > > Sent: Wednesday, June 19, 2002 1:33 PM
> > > To: 'Anoop Ghanwani'; 'stds-802-17@xxxxxxxx'
> > > Subject: RE: [RPRWG] SendA, SendB, SendC
> > >
> > >
> > > Anoop,
> > >
> > > We need 3 such signals because we have 3 separate service
> > > classes. Other 802
> > > MACs do not have distinct service classes. A client is free
> > > to ignore the
> > > signals if does not want to use them or does not know how
> > to use them.
> > >
> > > jl
> > >
> > > -----Original Message-----
> > > From: Anoop Ghanwani [mailto:anoop@xxxxxxxxxxxxxx]
> > > Sent: Wednesday, June 19, 2002 11:35 AM
> > > To: 'stds-802-17@xxxxxxxx'
> > > Subject: [RPRWG] SendA, SendB, SendC
> > >
> > >
> > >
> > >
> > > Why does 802.17 need SendA, SendB, SendC signals
> > > going from the MAC to the client? Other 802 MACs
> > > do not have a similar primitive even though they
> > > have to get the client to wait in order to get
> > > access to the medium (e.g., in token ring LANs, a
> > > station must wait until a free token arrives).
> > >
> > > Can anyone shed some light on this?
> > >
> > > -Anoop
> > > --
> > > Anoop Ghanwani - Lantern Communications - 408-521-6707
> > >
> >
>