Re: [Int-area] Intdir early review of draft-ietf-intarea-gue-06

[Tom Herbert <[e-mail]>]

Hi Charlie, thanks for the review!

On Thu, Feb 28, 2019 at 7:00 PM Charles Perkins <[e-mail]> wrote:
>
> Reviewer: Charles Perkins
> Review result: Almost Ready
>
> This document needs an applicability statement which includes the assumptions
> and the reasons it might be useful.  Deliverability needs to be expanded.

I will add an applicability statement.

> Reasons why middleboxes would be unlikely inspect GUE fields might be included.
>
See comment below.

> For instance, the discussion in second paragraph of 5.11.1 belongs in the
> applicability statement.
>
Okay

> It should also be explained why arbitrary GUE extensions are less likely to be
> filtered out compared to IPv6 destination options.

I'll clarify the requirements about middle boxes parsing and
processing GUE headers and make it clear MUST NOT to do this and a
SHOULD NOT for inspecting GUE header. However, as long as GUE is in
plaintext and sent to a well-known port number there's nothing we
could do to prevent middlebox inspection or this sort or filtering if
someone really wants to do it. In fact the draft assumes inspection
will happen and sets requirements for it. Authentication of the GUE
header would at least be a way to detect if middelboxes modify the
header.

> ============================================== The document assumes close
> familiarity with deployment scenarios that seem to be characterized by acronyms
> such as RSS, aRFS, TSO, LRO, etc.  While I am pretty familiar with a lot of
> encapsulation techniques, I had to study the meaning of these acronyms.  If it
> is intended to effectively restrict the intended audience, that is O.K., but
> otherwise more background is needed along with relevant citations.

I'll add some referecences, however note that this is in the appendix
(although Appendix A doesn't have the blurb about just being
informative so I'll add that).

> ============================================== [GUEEXTENS] is cited in a way
> that places a normative dependency on [GUEEXTENS].  So, [GUEEXTENS] belongs in

Okay.

> the Normative References. ============================================== I have
> a large number of specific comments which I will post shortly in the form of a
> rfcdiff-generated file.
>

Tom