Re: [sidr] I-D Action: draft-ietf-sidr-usecases-03.txt

["Murphy, Sandra" <[e-mail]>]

speaking as regular ol' member

I'm not really sure what you are talking about here.

You speak of bgpsec and signing - that is not origin validation, that is path validation.

Origin validation is strictly and only judging whether the origin has the authority to originate announcements of the prefix.  Origin validation does not ensure that the authorized AS is actually making the announcement.

It is bgpsec path validation that ensures that the announcement is currently and authentically being announced by the AS.  That's the "signed injection" you mention.

--Sandy

________________________________
From: Brian Dickson [[e-mail]]
Sent: Friday, March 30, 2012 8:55 AM
To: Murphy, Sandra
Cc: Christopher Morrow; Sriram, Kotikalapudi; Chris Morrow; [e-mail] list
Subject: Re: [sidr] I-D Action: draft-ietf-sidr-usecases-03.txt

Sandy,

The use case example I included, is an origin validation use case.

The problem is the inability to have "origination" (signed injection into BGPSEC) done by proxy, at all or especially by more than one party.

The particular situation leading to the need for this, would be a stub AS whose router vendor does not support BGPSEC, or whose router hardware or code base is not included in BGPSEC support by their router vendor.

(Doing a forklift upgrade to do BGPSEC is generally seen as #FAIL by operators.)

This is likely to be a fairly common situation _among stub AS_ and, given the current requirement for origin injection for BGPSEC, addressing this _may_ impact the use cases document, for _origin validation_.

I'm just saying...

Brian

On Fri, Mar 30, 2012 at 4:21 AM, Murphy, Sandra <[e-mail]<mailto:[e-mail]>> wrote:
Brian, Chris.

The usecases draft was intended to describe origin validation use cases.

Route leaks (and other path validation issues) might need their own usecases draft.

But I don't think we should add those cases to this draft.

--Sandy

________________________________________
From: [e-mail]<mailto:[e-mail]> [[e-mail]<mailto:[e-mail]>] on behalf of Christopher Morrow [[e-mail]<mailto:[e-mail]>]
Sent: Thursday, March 29, 2012 7:21 PM
To: Brian Dickson
Cc: Sriram, Kotikalapudi; Chris Morrow; Murphy, Sandra; Murphy, Sandra; [e-mail]<mailto:[e-mail]> list
Subject: Re: [sidr] I-D Action: draft-ietf-sidr-usecases-03.txt

On Thu, Mar 29, 2012 at 3:11 PM, Brian Dickson
<[e-mail]<mailto:[e-mail]>> wrote:

> I think the use cases are likely to be informed by protocol design, so even

s/informed by protocol design/altered if the protocol design changes/

I'm not sure if the protocol design's going to change the use-cases...
you're still going to want to secure a route. (not an important point)

> I have a few examples that I can think of, which would necessarily depend
<snip>
> I'd prefer this to be added to a "raft" of IDs, for which there is no rush
> to publish until they are all completed, after which the timing would be
> appropriate.

I'm not against this, though we've got a document hanging out post
WGLC (perhaps it ought to be re-reviewed if the changes were
significant... anyway) and we'll have to keep kicking it each 5.5
months to stay 'alive'. (again, not super important, and see below as
well)

> Here's an example of use-case, which depends on certain assumptions (which
> may or may not be appropriate, but which are fodder for discussion):
>
> Suppose there is an Edge-AS "E", and transit providers to "E", which would
> be "X" and "Y".
>
> Suppose "E" does not do BGPSEC (per se), but wants to have BGPSEC signing
> done "for her", by "X" and "Y".
>
> (Ignore for the moment that the _current_ designs don't support that, that
> is an entirely other rat-hole for the moment.)

hrm, in:

 <http://tools.ietf.org/id/draft-ietf-sidr-bgpsec-ops-04.txt>

section-6 there's discussion of 'only sign your one prefix, do nothing
else complex' which fits the model, albeit requiring the end site to
run some small number of commands on their device. If they wanted to
hand their private key materials to the upstreams they could do the
signing, but that seems icky (to me).

I don't know that, if implications are understood by the end site and
configurations available for use on their side, end-sites would want
to hand over control of their IP assets in this way. Running the
signing on their side should be simple enough, and low/no-cost.

> And publishing something IMHO prematurely, locks the WG into that RFC,
> making revising it much harder, than if it were still in-WG and
> not-yet-published.

I think the authors said something like: "send text" where you think
it is fit to be inserted... If other folks want to delay/re-review
they need to speak up. Consensus so far was that the document was
ready to move along.

-chris