Hi Pekka,

Thanks for the fast feedback.  I'm sorry for my long delay in responding.

On 3/18/07, Pekka Savola <[EMAIL PROTECTED]> wrote:

On Sun, 18 Mar 2007, Alia Atlas wrote:
> For a slightly different approach at the problem, take a look at
> draft-atlas-icmp-unnumbered-02.txt. It has the ability to report on
> what would have been the outgoing interface, as well as the incoming
> interface.  The information can include an ifIndex, an IP address,
> and a brief string description.
>
> I'd be very interested in your feedback.

The draft is unclear wrt 'description'.  It says that it SHOULD report
'ifName' but other info can be reported as well. This is very
confusing as 'description' has established meaning in the ops
community and that meaning is NOT ifName.  The object should be named
'Interface name' (ifName) because that's distinct from description
(ifAlias).


Yes, I'll change it to interface name now.

I'd very strongly object to reporting 'description' (ifAlias).  In
deployments I've seen this often includes confidential information
(e.g., customer numbers, link identification numbers, etc.).  I
suggest rephrasing this so that the name of the object is 'Interface
name' (ifName).  If description (ifAlias) would be retained, the spec
should require that by default it MUST NOT be reported.


Yes, I agree and have changed it to be interface name, and clearly ifName
as a SHOULD.

In general, I find the specification having features that I don't see
necessary. Reporting ifIndex is not necessary (and the index is not
very useful as is to a human) if ifName is reported.  Addresses are
already known from the source address though somewhat less reliably so
those need not be reported for outgoing interface use, and could also
result in reporting IPv6 link-local addresses (or IPv4 private
addresses) which wouldn't necessarily be useful or desired.


The idea with reporting the ifIndex is to provide the easy ability to
correlate
that to MIB data.  It is a common look-up key and I believe it to be useful.

Using simply the source address doesn't handle topologies with parallel
links between routers.  In that case, knowing the exact outgoing link for
troubleshooting
is useful.  I believe that even IPv6 link-local addresses & IPv4 private
addresses
would be useful in distinguishing between two interfaces; however, if that
isn't desired,
it would be under operator control.

There also seems to be some specification about incoming/outgoing
interface (which is probably why addresses are supported).  This is
distinct from what draft-watkins-icmptype11code0 is about because that
draft reports next-hops which this doesn't do.


Yes, I'm primarily interested in the incoming interface, for better
traceroute
troubleshooting support, but others were interested in the outgoing
interface.
If there is a desire to include the outgoing next-hops, that could be added
as
a sub-object.

Personally, I'd prefer seeing either one big document or two
documents, one describing 'ifName' reporting and one describing IP
address reporting (nexthop, possibly input/output interface).
Currently input/output address reporting seems to be out of scope of
this draft though: the title, draft name, abstract, etc. don't reflect
these address extensions (meant to identify the incoming interface) so
some rewording at the very least is needed there.


Well, my initial focus for the draft was on handling unnumbered interfaces,
but other authors were also interested in the IP addresses.  The abstract
does address appending the IP address to facilitate troubleshooting; this
is primarily using the incoming interface.

I'd be happy to add another sub-object if that addresses the concerns of
obtaining the next-hop and that is desirable.

As a more editorial issue, the first two paragraphs could maybe use
slight modification to better account for IPv6 (e.g., ref to ICMPv6
spec).  As a detail, source address selection for ICMPv6 is specified
slightly differently, but it's spec-compliant to do it the same way as
for IPv4.  It's not guaranteed though.


I'll add the ref to the ICMPv6 spec.

Thanks for the feedback,
Alia

--
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings

_______________________________________________
Int-area mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/int-area

Reply via email to