Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [RPRWG] Gandalf - question on Framing



Hi Siamack,

I was not meaning uniquness but the fact that we are defining a new MAC 
layer.

We are, in principle, free to place the RPR header fields in any order 
we like (as a joke why not DA, TTL, SA, CoS, PT, ..., HEC?).

However, there is a criteria that imposes us to allow a simple mapping 
of RPR frames in Ethernet frames, thus it is a good idea to put the 
three fields that are common with Ethernet (i.e. DA, SA, PT) in the 
same order as Ethernet.

The two competing ideas arguing whether to put the RPR specific fields 
before or after the Ethernet-like fields are both techincally sound. 
The discussion on this issue is more political (e.g. reusing the 
existing testing equipment and/or off-the-shelf Ethernet MAC chips).

What I am arguing, from the technical perspective, is that we do not 
need a field in the RPR header that says that the frame is an RPR 
frame: it is an RPR frame by definition.

We would have needed such a field if we designed RPR as a ring-aware 
switching technology, built on top of point-to-point Ethernet MACs. But 
this is not actually what we did.

Regards, Italo

> -----Original Message-----
> From: sayandeh@xxxxxxxxxx [mailto:sayandeh@xxxxxxxxxx]
> Sent: Wednesday, December 05, 2001 5:32 PM
> To: Italo.Busi@xxxxxxxxxx
> Cc: sayandeh@xxxxxxxxxx; tak@xxxxxxxxx; pkj@xxxxxxxxxxx;
> stds-802-17@xxxxxxxx
> Subject: Re: [RPRWG] Gandalf - question on Framing
> 
> 
> 
> 
> Italo.Busi@xxxxxxxxxx wrote:
> 
> > Having the Ethernet based frame with the Type field saying 
> that we are
> > carrying RPR means, IMO, that we are building a 
> ring-oriented switching
> > technology on top of point-to-point Ethernet MACs.
> > This is not what we are doing according to our 5 criterias 
> and PAR, as
> > Mike said. Both Allading and Gandalf are aligned on this issue.
> 
> RPR also uses a point to point physical layer to build rings. 
> Its difference
> to an Ethernet MAC includes fairness, resiliency, topology 
> discovery, ring
> selection, new client interface, etc. This should be enough 
> for uniqueness.
> 
> Regards, Siamack
> 

WINMAIL.DAT