In an attempt to simplify my earlier example, I oversimplified
it. Here is a better diagram:
>The ICMP issue that you have described would also exist when using
>multiple "conceptual" routing tables in the case of a partitioned
>site. For example:
==========================================================
SITEA
Host1 Host2
| |
______._|_________ ___.______._|_____________
Link1 | | | Link2
| (down) |
| | |
|+-----------------+ |
|| |
============R1========================R3===================
SITEB | |
| |
_____|_________________________|________________
Link3
===========================================================
Now, if Host1 sends a packet with site-local source and
site-local destination, it obviously won't reach Host2.
R1 will receive the packet, and will be unable to route
the packet because there are no routes available in the
conceptual site-local routing table for SITEA. This
will result in an ICMP/No Route to Destination error.
The use of the conceptual routing table prevents R1 from
noticing that there is a global route to Link2, so the
potentially useful ICMP/Scope Exceeded error is not sent.
The user may see that some applications can reach Host2
and other can not, but no pertinent error message will
be displayed. Also, "smart" applications do not receive
the information that addresses of a greater scope
could potentially reach the destination.
But, if Host1 sends a packet with site-local source and
global destination through the same path, the ICMP/Scope
Exceeded message will be returned.
Is it is important to maintain the "Scope Exceeded" message
_only_ for packets where the destination scope is greater
than the source scope? Even if it means that we may discard
packets that could successfully reach the destination in
some cases (as in my first example)?
A mixed scope packet where the source is greater than the
destination is the _only_ time that a Scope Exceeded message
will be generated in the current architecture. And, it
is one of the least useful times, since the mixed scope
packet would not have been generated if the sender had the
option of using a matching scope address -- either because
a global address is not available, or because an upper layer
has chosen the addresses. Also, the mixed scope packet
should be fairly obvious to a sophisticated network debugger,
even if it were returned in an ICMP/No Route to Destination
message.
Margaret
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
