Re: WG last calls

["Adrian Farrel" <adrian@olddog.co.uk>]

Hi,

Yes, to clarify what Jerry says, the number of crankback attempts MAY be
limited. At the moment, the only way that we provide to limit the attempts
is implemented per LSR. That is, any LSR MAY decide that it has performed
enough attempts at rerouting and pass the error back to the upstream LSR.

Note that the use of crankback to derive a path through a network is not
recommended. This approach is almost equivalent to random walk routing and
is neither efficient nor effective.

The use of crankback, as described in the draft is intended for use in
specific circumstances, such as inter-domain routing. In these cases only
selective LSRs (such as domain boundaries) perform rerouting attempts.

Adrian

----- Original Message ----- 
From: "Ash, Gerald R (Jerry), ALABS" <gash@att.com>
To: "Zhang Renhai" <zhangrenhai@huawei.com>; "Adrian Farrel"
<adrian@olddog.co.uk>; <ccamp@ops.ietf.org>; "Kireeti Kompella"
<kireeti@juniper.net>
Cc: "Ash, Gerald R (Jerry), ALABS" <gash@att.com>
Sent: Friday, April 01, 2005 2:02 PM
Subject: RE: WG last calls


Zhang,

> From draft-ietf-ccamp-crankback-04.txt ,section 4.5, the
> number of crankback rerouting is limited, so there will
> be a result: there is another LSP,but for the reason of
> limiting, the path may be not found.  I think sometimes
> it is unacceptable.The LSP may be prefered to enhancing
> performance.

This problem can be avoided by the setting the node retry threshold
(configurable per node) very high (~infinity), so retries aren't
limited.

Thanks,
Jerry