Re: Single subnet

Steve Deering <deering@pescadero.stanford.edu> Sun, 20 May 1990 00:04 UTC

Received: from merit.edu by NRI.NRI.Reston.VA.US id aa09023; 19 May 90 20:04 EDT
Received: Sat, 19 May 90 19:03:43 EST from Pescadero.Stanford.EDU by merit.edu (5.59/1.6)
Received: by Pescadero.Stanford.EDU (5.59/25-eef) id AA29129; Sat, 19 May 90 17:03:30 PDT
Date: 19 May 1990 16:20-PDT
From: Steve Deering <deering@pescadero.stanford.edu>
Subject: Re: Single subnet
To: Dave Katz <katz@merit.edu>
Cc: fddi@merit.edu
Message-Id: <90/05/19 1620.069@pescadero.stanford.edu>
In-Reply-To: Dave Katz's message of Sat, 19 May 90 180251 EST
Status: O

> >  Will CLNP routers detect the wrap->thru transition and use
> >  that to trigger a solicitation for ES Hellos?
>
> Yes, in that IS-IS uses the loss of an IS on a subnetwork to trigger
> this function.  

If all the ISs are dual-MACed, unwrapping does not cause any ISs to
disappear from either of the two resulting subnetworks.  So, for this
to work, it would have to be the loss of an IS *at a specific MAC address*
that triggers the solicitation of ES configuration.  Is that the case?

Also, if there is only one (dual-MAC) IS attached to the FDDI network,
it would have to detect the loss of connectivity between its *own* two
MACs, and use that to trigger ES solicitation.  Is that the case?

I presume that the FDDI MAC or PHY layer detects unwrap events.  Can they
or should they not be signalled upward as "change of topology" events?

> However, asking for ES Hellos doesn't help get rid
> of the ones you can't reach any more.  The spec calls for reducing
> the remaining holding time on ES adjacencies, but when you add
> it all up, it can take up to 239 (!) seconds to lose old ES adjacencies.

But that's the main problem, right?  An IS may continue to send packets
to a single-MAC ES via the "wrong" ring for that length of time after an
unwrap event.  (I suppose a fancy IS would load split across the two
rings, so half the packets would get through.  Are the transport
protocols expected to cope with that?)

> These timers can be cranked down, however.

Sounds like a good idea!

Steve