Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02

Marc Petit-Huguenin <petithug@acm.org> Mon, 10 August 2009 16:52 UTC

Message-ID: <4A805049.7020709@acm.org>
Date: Mon, 10 Aug 2009 09:52:25 -0700
From: Marc Petit-Huguenin <petithug@acm.org>
User-Agent: Mozilla-Thunderbird 2.0.0.22 (X11/20090701)
MIME-Version: 1.0
To: Dave Thaler <dthaler@microsoft.com>
References: <0a9801ca10f2$8f933600$5f7d150a@cisco.com> <E4561B14EE2A3E4E9D478EBFB5416E1B196755@TK5EX14MBXW653.wingroup.windeploy.ntdev.microsoft.com>
In-Reply-To: <E4561B14EE2A3E4E9D478EBFB5416E1B196755@TK5EX14MBXW653.wingroup.windeploy.ntdev.microsoft.com>
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7bit
Cc: "draft-ietf-behave-turn-uri@tools.ietf.org" <draft-ietf-behave-turn-uri@tools.ietf.org>, "behave@ietf.org" <behave@ietf.org>
Subject: Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02
Precedence: list

Thanks for your review.  Please see my answers below.

Dave Thaler wrote:
> Attached is a copy with my comments.
> 
> Summary of comments that aren't purely editorial:
> 
> 1) Why does this diverge from the generic URI syntax in
>    RFC 3986 by omitting // before the host part?

I went back and forth between a hierarchical URI (i.e. with the //)
and and opaque URI (without the //), but the changelog[1] that
explain this disappeared between the draft-petithuguenin and the
draft-ietf versions (I put back the complete changelog in my local
copy of the draft, to be publish at the end of the WGLC).

The reason I went back to an opaque URI is because of RFC 4395
Section 2.2.  The TURN URI does not describe a hierarchical
structure and RFC 4395 advises to use opaque URI in this case.

> 
> 2) The algorithm keep saying convert to "a" TURN transport
>    "as specified above".  However, there's nowhere in the doc
>    that actually does so.  

The "as specified above" refers to the fourth paragraph of section 4
("In some steps...").  I inverted "<transport>" and "<scheme>" in
the fourth paragraph so it matches the order used in the steps.  I
did not find a better way to do not have to repeat the text multiple
times.  Any advice?

> Page 4 _filters_ the list, but
>    you can end up with multiple (turn/tcp -> TCP and TLS).

I am not sure to understand this.  The conversion described is for
the URI elements i.e.:

turn:example.com?transport=udp --> UDP
turn:example.com?transport=tcp --> TCP
turns:example.com?transport=tcp --> TLS

There cannot be multiple conversions at this step.  Do you have any
suggestion to improve the text?

> 
> 3) Is there any ordering constraint on the order in which
>    IP addresses are tried or is RFC 3484 sufficient?

I think that RFC 3484 is sufficient.  Should I add a reference?

> 
> 4) Algorithm step 3 is underspecified.  It says it recommends
>    "A or AAAA query".  In constrast step 2 says "A and AAAA
>    query".  Why is step 3 an "or"?  I think it should be
>    "and".

Yes, modified on my local copy.

I also applied all your editorial modifications.  I moved the
"Running Code Considerations" section in the "Release Notes"
appendix so it will be removed (with the URIs section) before
publication (see draft-petithuguenin-running-code-considerations for
the motivations and the subsequent discussion on the IETF mailing
list on how much people disliked it :-( )

[1]
http://tools.ietf.org/rfcdiff?url2=draft-petithuguenin-behave-turn-uri-03.txt

-- 
Marc Petit-Huguenin
Personal email: marc@petit-huguenin.org
Professional email: petithug@acm.org
Blog: http://blog.marc.petit-huguenin.org

[BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Dan Wing
Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Dave Thaler
Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Marc Petit-Huguenin
Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Alfred E. Heggestad
Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Dave Thaler
Re: [BEHAVE] WGLC: draft-ietf-behave-turn-uri-02 Marc Petit-Huguenin