I am happy with the text that addresses my DISCUSS, but I
suggest that IEEE 802.3, IEEE 802.1Q and IEEE802.1ad be included as Normative
References.
Dan
From: Maria A. Dos Santos (mariados) [mailto:[EMAIL PROTECTED]
Sent: Friday, August 04, 2006 10:03 PM
To: [EMAIL PROTECTED]; [email protected]; [EMAIL PROTECTED]; [EMAIL PROTECTED]; Jari Arkko; Cullen Jennings (fluffy); Mark Townsley (townsley); Romascanu, Dan (Dan)
Subject: Latest updates : draft-ietf-l2tpext-pwe3-ethernet-08.txtHi all,The following is the consolidated list of all the LC comments and responsesby different people. I highlighted the responses and what/how will be replaced.This has been reviewed by Ignacio Goyret once already and I alreadyincorporated his last batch of editorial change requests in the attachedversion 8 of the draft.There is still one comment from Cullen Jennings which is not resolvedyet, but no response from Cullen yet. Do others want to advice on howto handle this comment?regards,Alice-------------------------------------------------------------------------------------------------------------
###
Comments raised by Francis Dupont, with additional comments/
responses inserted by Brian Carpenter, Mark Townsley and
Ignacio Goyret.- Abstract: This document describes transport -> the transport?Fixed- 1.1: remove trailing "." of PE definitionFixed- 1.1: I suggest to introduce here the L2TPv3 messages as
imported abbreviations in the logical (i.e., not alpha) order.Added section 1.2 to introduce the l2tp message types- 1.2 (NSP function): Ethernet packets -> Ethernet framesFixed- 1.2: as VLAN tags are in some further cases 802.1q tags,
I propose to cite 802.1Q the first time and use "VLAN" everywhere
at the exception of 802.1Q specific places.Inserted elaboration for terms such as "Ethernet" and "VLAN" in
Introduction section.- 2.2 c): must -> MUST?Fixed- 2.3.2: to indicate the Ethernet interface -> plural without "the"?Fixed- 3.2: may -> MAY?Fixed- (technical) 3.3: I am not convinced at all by the very last part
(fragmentation/reassemble recommendation) because it is a layer violation
and the goal (manage the issue at only one place as explain in
L2TPFRAG section 5.1) is not explained.Replacement suggested by Mark Townsley."As a result, the pseudowire endpoints SHOULD support
Fragmentation and Reassembly procedures specified in Section 5 of
[L2TPFRAG] or section 4.1.4 of [RFC3931] when the Ethernet payload
is IPv4. Only one of these methods SHOULD be used for a given
Ethernet PW."has been changed to:This is not specific only to Ethernet over L2TPv3,
and the base L2TPv3 specification [RFC3931] provides general
recommendations with respect to fragmentation and reassembly in
section 4.1.4. "PWE3 Fragmentation and Reassembly" [L2TPFRAG]
expounds on this topic further, including a fragmentation and
reassembly mechanism within L2TP itself in the event that no other
option is available. Implementations MUST follow these guidelines
with respect to Fragmentation and reassembly.
- 4: please introduce the AC abbrevFixed- 4: the 802.1Q tag -> a VLAN tagFixed- 4: preamble or FCS -> preamble and FCSFixed- 4: IPSec -> IPsec (cf RFC 4301 introduction)Fixed- 4: For Ethernet VLAN PW, VLAN tag rewrite ...: can't parse this statement
Replacement proposed by Mark Townsley:While architecturally [RFC3985] outside the scope of the L2TPv3 PW
itself, if VLAN tags are present, the NSP may rewrite VLAN tags on
ingress or egress from the PW (see section 3.1)."Replacement proposed by Brian Carpenter:The VLAN tags of an Ethernet VLAN PW may be rewritten by NSP at
the egress LCCE, which is outside the scope of this document (see
Section 3.1).Mark's suggestion is incorporated into draft.
- 4: there is no TOS octet or QoS field in the IP header (look at RFC 2474?)
Mark Townsley:> No doubt. We need to at a minimum change this to "ToS bits"and strike
> "IP QoS field" - the paragraph should perhaps simply be rewritten to
> refer to DSCP only.
Exchanges between Brian and Ignacio:>>>[BC] Indeed, there is no "for example" about it; there only the DSCP.
>>
>>
>> RFC2474 specifically refers to that byte as "the TOS octet" (eg, see
>> the second to last paragraph on the abstract, page 2).
>>
>> How about eliminating the parenthesized "for example, according to DSCP"?
>> Would that be sufficient?
>
>Well, it is called the TOS octet in IPv4 and the Traffic Class octet in
>IPv6 - the only phrase that applies to both is DSCP. So either you have
>to put "TOS or Traffic Class octet" or refer to 2474, I think.
Changed reference to "DS field" and added reference to RFC2474.
- 4: 802.1Q CS and ... -> too many "and" (note that here 802.1Q is the
right term)Fixed5: should -> SHOULD ??From Mark:I'm not entirely convinced that this is a SHOULD given that it isn't
really a protocol mechanism that we are defining here, but a
deployment recommendation.So this stays unchanged.
9.1 L2TPFRAG: missing I-D name (draft-ietf-pwe3-fragmentation-10.txt?)Fixed
###########################################################################
*Lars Eggert:*
*Discuss:*
*[2006-06-30]* Section 5., paragraph 3:> LCCEs SHOULD monitor for congestion (by using explicit congestion
> notification, or by measuring packet loss) in order to ensure that
> the service using the Ethernet or Ethernet VLAN PW may be maintained.
> When severe congestion is detected (for example when enabling
> Sequencing and detecting that the packet loss is higher than a
> threshold) the Ethernet or Ethernet VLAN PW SHOULD be halted by
> tearing down the L2TP session via a CDN message. The PW may be
> restarted by manual intervention, or by automatic means after an
> appropriate waiting time. Note that the thresholds and time periods
> for shutdown and possible automatic recovery need to be carefully
> configured. This is necessary to avoid loss of service due to
> temporary congestion, and to prevent oscillation between the
> congested and halted states.RFC3985 also says: "Where congestion is avoided by shutting down a PW,
a suitable mechanism must be provided to prevent it from immediately
returning to service and causing a series of congestion pulses." I
don't see such a mechanism in this draft, only the above
acknowledgment of the issue and the recommendation of careful
configuration.(This is identical to the DISCUSS I have raised
against draft-ietf-pwe3-atm-encap. As far as I know, PWE3 is working
on an approach for addressing this. This document, however, belongs to
a different WG, so I'm not sure if any resolution that PWE3 comes up
with would apply to this document as well. I'd hope that it would -
can the authors/ADs confirm?)Confirmed by Stewart Bryant and Mark Townsley that the PWE3 congestion work
would be general and applicable to both L2TP and
MPLS based Pseudowires.
###########################################################################
*Cullen Jennings:*
*Discuss:*
*[2006-07-05]* I may not understand the congestion control technique but
it looks like it says, "in case of congestion, pull the plug". I don't
believe anyone will implement this congestion control suggestions - and
if they do, I don't believe the resulting systems will be usable - they
will be impossible to manage not to mention the incredible DOS
opportunities. No advice on the threshold is provided. I think this
needs to be updated to either be clear no congestion control is provided.From what I understand, the draft is inline with the current state of congestion
control for emulation technologies, and PWE3 WG is currently conducting
further study on the topic. Once we get the result of that, this draft
and all other pwe drafts (both l2tp and mpls) will leverage from that
mechanism. So at this point, the goal of the "Congestion Control" section
in this draft is only to emphasize the possibility of congestion and provide
some general best-effort recovery mechanisms. Would it be better if I emphasize
the goal?I am not sure what general default threshold is appropriate given that the threshold willvary according to network size and average traffic load of individual network. That's whyit is left undefined and left to administrator to experiment.The section titled "Applicability Statement" gives me no idea where this
is applicable and where it is not.Would you mind elaborating what information are you expecting here?###########################################################################
*Dan Romascanu:*
*Discuss:*
*[2006-07-03]* I made the following two comments during the IESG LC. The
comments were not addressed:1. The I-D speaks about 'Ethernet' but does not provide any reference.
What version of the Ethernet standard is targeted here? The latest
approved version of the Ethernet standard is IEEE Std 802.3-2005, but
the IEEE 802.3 WG also has in works an extension of the frame size as
IEEE 802.3as. One cannot talk about Ethernet PDU and MTU without a
proper reference.2. The I-D speaks about 'Ethernet VLAN'. Strictly speaking there is no
such thing, Virtual LANs are being defined by IEEE 802.1 and not only
for Ethernet. However accepting terminology license based on the
de-facto reality that most if not all VLANs run over Ethernet, there
still is a need to specify if the document refers to IEEE 802.1Q VLANs
(one 12-bit tag) or IEEE 802.1ad which accommodate multiple tags and
introduce the concepts of Customer VLAN and Service VLAN. Maybe the
VLANs referred in this I-D are transparent to the type of VLANs that are
being connected, but even so there should be text explaining this.Elaboration for the term "Ethernet" and "VLAN" suggested by
Ignacio Goyret and has been inserted into Introduction section:" The term "Ethernet" in this draft is used with the intention to
include all such protocols that are reasonably similar in their
packet format to IEEE 802.3, including variants or extensions which
may or may not necessarily be sanctioned by IEEE (including such
things as jumbo frames, etc). The term "VLAN" in this draft is used
with the intention to include all virtual LAN tagging protocols such
as IEEE 802.1Q, 802.1ad, etc."
_______________________________________________ Gen-art mailing list [email protected] https://www1.ietf.org/mailman/listinfo/gen-art
