On Aug 30, 2007, at 12:31 AM, Arnaud Ebalard wrote:
Dow Street <[EMAIL PROTECTED]> writes:
IP seems to be the right place in the stack to deal with issues of
path selection and packet delivery. Do you see a better place in a
different layer?
If you need something working at a large scale, which creates a
_precise_ rationale to have that included at L3, it is the place and
SHOULD be put there.
Now, if there is only a limited rationale, it COULD be put there
but as
an external feature (i.e. outside 2460), to be deployed by people that
have the need.
<snip>
Can you clarify what you mean by "blindly"?
What 2460 specifies for routers regarding RH0 processing and what
vendors had implemented (before the deprecation). I don't consider the
amplification here (i.e. even a limited version of RH0 with only one
address in it, processed by default by routers, fits my description).
Are you primarily concerned with making / keeping source routing a
*mandatory* component of IPv6 because you see this as adding attack
vectors at every IPv6 node?
I suppose I see these as separable:
- making the functionality mandatory is to help get it into
implementations so that interested users / operators have source
routing as a tool.
- this does not mean (to me) that the functionality to act as an
intermediate source routing destination (routers and/or hosts) must
be enabled by default. I think this question is worth discussing.
It probably also makes sense to consider architectural and
implementation issues separately:
Architectural: I think this is our primary area to address in any
new RHx draft. As I understand it, RH4 directly addresses the
architectural issue of RH0 that was of most concern to the group -
the number of intermediate way-points.
Implementation: A new RHx draft can probably provide better
implementation guidance than was present for RH0, helping to reduce
the number of implementation problems. I don't think concern over
implementation problems is a valid reason for eliminating source
routing altogether.
There also appears to be a wider philosophical debate going on (among
list participants) over general vs. specific functionality, and the
type of network we think the Internet should be. Should the Internet
be a collection of specific mechanisms, each oriented toward a known,
narrow, and demonstrably profitable use? Or should the Internet
architecture (whatever that is) provide a set of general tools that
can be used in a variety of ways? As IETF standards go, you don't
get much more "architectural" than IP.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------