TSVWG
Allison Mankin
James Polk
Jon Peterson
Chaired by Allison & James – Jon is in ENUM
TSVWG Session I (1 hour), 10 minute break, Session II (1 hour)
TUESDAY, November 8, 2005
1740-1840 Afternoon Session III
1850-1950 Afternoon Session IV
-------------------------------------
1) Agenda Bashing 5
Note Well
Introduction of the "Experiment" (i.e. James Polk as new
WG Chair)
Work through break?
Decided we will work through break, and allow anyone to get
refreshments as wanted.
===================================================================
Fred Baker:
What happens to the mlpp-that-works draft?
Allison Mankin:
We will discuss this in the re-chartering section of the agenda.
2) Bob Briscoe
draft-briscoe-tsvwg-re-ecn-tcp-00 20
===================================================================
Adding Accountability for Causing Congestion
draft-briscoe-tsvwg-re-ecn-tcp-00.txt
Originally targeted for IPv6, reshaped it to IPv4 for this draft.
Intended for standards track.
Main concern:
Non-compliance with e2e congestion control (e.g. TCP-friendly)?
Not just per flow congestion response
-- smaller: per packet
-- bigger: per user
-- even bigger: per up stream network
Previous Work
-------------
- detect high absolute rate [commercial boxes]
- sampled rate response to local congestion [RED+ sin bin]
- transport control embedded in networks [ATM]
- honest senders police feedback [nonce]
Basic Idea
----------
- Sender re-inserts congestion feedback into forward data:
"re-feedback"
-- On every Echo-CE, mark ECT(0) else mark ECT(1).
Talks about the notion of Credit and Debit based on the reference point
of balance (difference) of ECT(0) and CE marked packets at any point in
the path. This allows the notion of upstream congestion and downstream
congestion from the reference point where this balance is tested.
The goal of the balance is CE marked packets = ECT(0) marked packets.
Talking about the Egress Policer and the Ingress Policer.
Notice for Ingress Policer, depends on RTT.
For RTT need sister proposal for "re-TTL". Currently not in this
draft.
Accountability for Congestion Applications:
------------------------------------------
- congestion-history-based-policer (congestion cap)
- DDoS mitigation
- QoS & DCCP profile flexibility
- Load sharing, traffic engineering
- Bulk metric for inter-domain SLAs or charges
Requires Feedback Established (FE) flag in IPv4 or IPv6
header/extension.
Bit 48 in the IP header is currently un-used, thinking of using this
for FE flag.
RE-ECN Limitations
------------------
- This relies on ECN being used.
- Dependency on getting re-TTL standardized.
- Takes a while for dropper & policer to detect malice
-- Dynamic attacks not detected fast enough.
- Flow state at ingress policer and egress dropper.
Summary
-------
- Accountability has been a weakness of the Internet so far.
- Request that ECN nonce be held as experimental
Next Steps
----------
- finish RE-ECN draft
- do RE-TTL draft
Fred Baker:
On slide 11 of the inter-provider picture.
Indicating that congestion not expected between providers.
But at the link right before the receiver. Does this solve
the attack by causing the receiver to pay?
Matt Mathis:
Question on if sitting behind a slow speed modem.
Does this method end up charging a user using a slow link
sending small amount of traffic?
Sally Floyd:
We are not using a DSCP to separate this from current ECN nonce?
Bob Briscoe: Correct.
Sally: This will hold up the current nonce RFC.
(Joe Elally ??):
Can we be dropping packets that has already paid?
Bob Briscoe:
If interest in (Good Put??), this will have the correct incentive.
If interest in DOS attack, the info can be sent upstream to
stop the flow upstream to help mitigate the attack.
3) Sally Floyd
draft-floyd-ecn-alternates-02.txt 5
draft-ietf-tsvwg-quickstart-01.txt 5
draft-kuzmanovic-ecn-syn-00.txt 5
===================================================================
Alternate semantics of ECN draft
--------------------------------
- Should this be BCP or Informational?
Changes from -01
----------------
3 changes, please see slide.
Georgios Karagiannis:
End to End ECN and Edge to Edge ECN differences was not called out in
the draft.
Sally:
Both will need to satisfy this draft.
Please send any suggestions on this.
Sally:
goal to be a BCP, wants WGLC.
Hum for submission as WGLC.
To be submitted as BCP.
draft-ietf-tsvwg-quickstart-01.txt
----------------------------------
Changes from last IETF: (See slide)
----------------------
- added 30-bit QS Nonce.
- changed IP tunnels
To Do:
-----
Delete the sentence in Section 4.6.2
Response feedback from Bob Briscoe
Talks about the 30 bit QS Nonce (see slide)
Changed heavily on section on IP Tunnel and IPsec AH
based on comments from David Black and Joe Touch. (see slide)
Joe Touch:
The tunnels that break are the ones at the receiver end and they
are not detectable.
Feedback from Bob Briscoe slide (see slide).
Bob Briscoe: Asks to see what Bob's new nonce will do to this.
(XXX ??):
Can it use the TCP initial sequence number for this?
Sally:
Need a number that the receiver cannot get/guess.
Need a random number, hence TCP initial sequence number will not work.
Bob Briscoe:
Clarifying that receivers are also senders, hence how
can one that does not trust receiver trust senders? when
a node plays both roles?
No group action at this time on this ID.
Presentation on draft-kuzmanovic-ecn-syn-00.txt
-----------------------------------------------
Changes to 3168 to allow TCP SYN/ACK be ECN capable.
Can this be made as WG work item.
The hum indicates this should be WG item.
Sally to resubmit this as WG item.
*chair NOTE - since the meeting, this ID is in chair discussions to
move to the TCPM WG, if they would take this effort, as this has more
to do with TCP then ECN. Chairs will let the WG know if/when status
changes to this ID.
8) Chairs / Charter Revision 30/remainder
===================================================================
Chairs moved discussion up to make sure enough of the WG was present
for this discussion
- This WG will state that it is where maintenance for RSVP and SCTP
will take place.
The existing Charter's Milestones are very out of date
------------------------------------------------------
- Early Retransmission to IESG.
- Submit SCTP Implementer's Guide on IESG tele-chat for Dec 1.
- 2009 milestone date...
New Milestones
--------------
See slides
Asking for dates for SCTP docs.
WGLC after this meeting:
- Implementing an Emergency Telecommunications Service for Real Time
Services in the Internet Protocol Suite
- RSVP extension for the Reduction of Bandwidth of a Reservation flow
- DS Service Classes
- DSTE (but will require 3 expert reviews posted to list prior to WGLC
ending)
Sally:
Can add Alternate ECN to Dec 05 batch?
Subha Dhesikan:
should also add nested VPN to the WGLC list.
RSVP IPsec? Under Security Area review.
On the mic (??):
Behave WG looking at SCTP, may want to consider their comments wrt
SCTP.
ECN Nonce to PS, need to discuss this offline with Bob's new work.
TCP Friendly Variable Rate Control to DCCP, Sally indicated yes on
this.
7) Michael Tuexen 10
draft-ietf-tsvwg-sctp-auth-01.txt
draft-stewart-tsvwg-sctpthreat-04.txt
addip status
===================================================================
SCTP Implementer's Guide (see slide)
----------------------------------
v-16 considered done by WG, ready for IESG review
SCTP Threat Doc
---------------
- discuss changes to the doc (see slide)
- Should this become a WG item as Informational
-- 50% of room hum to support as WG item, no hum opposing.
-- WG doc for next rev.
SCTP Authentication Doc
-----------------------
SCTP Add-In
-----------
- Talk about the handling of wild-card addresses.
SCTP Milestones (reflects date changes discussed during the meeting)
---------------
- I-G - Dec 05 (Informational)
- Add-IP and Auth - Feb 06 (PS)
- 2960BIS - Apr 06 (PS)
- Threats - Feb 06 (Info)
- API - June 06 (Info)
Allison asking for Temporary AD for handling the RFCs.
4) Philip Eardley 20
Francois Le Faucheur
draft-briscoe-tsvwg-cl-architecture-01.txt
draft-lefaucheur-rsvp-ecn-00.txt
===================================================================
Phil going over his slides on the Framework.
Georgios Karagiannis (At the flow diagram (slide 4)):
What happens if the source also sends ECN marking?
Phil:
If that is the case, we may use tunnels.
Bob:
There is admission control into this Diffserv class, hence
that will normally not happen.
Georgio:
Will follow-up with this on the list.
Phil explaining the Preemption mechanism.
Phil asking for feedback on this.
(Tom Phelan ??):
This is very similar to the RMD draft.
Why do it here, why not use RMD?
Phil:
We encourage the RMD group to work with us on this, at least
cross review the docs.
Georgios Karagiannis:
This is overlap with RMD.
Francois Le Faucheur:
We should try to see if we can converge.
Ted Faber:
Why are these in 2 different groups?
Bob Briscoe:
Thinks that ECN changes should be here, not NSIS.
NSIS should be working on signaling, not ECN.
But I may be wrong.
Ted Faber:
These groups should be in the same room.
(Tom Phelan ??):
Similar comments of using RMD draft.
Conclusion: We need to converge.
5) Francois Le Faucheur 10
draft-tsvwg-rsvp-dste-00.txt
draft-lefaucheur-rsvp-ipsec-02.txt
draft-lefaucheur-emergency-rsvp-00.txt
===================================================================
RSVP-DSTE-00 draft
------------------
Fred Baker, Kwok Ho Chan, and Subha Dhesikan volunteer as
WG reviewers of this ID, to be done during WGLC.
RSVP-IPSEC-02 draft (see slides)
-------------------
Changes 01 -> 02.
Handling dynamic SPI/Security_Association updates
- Need Security review on this.
Next Steps
- Complete Security Area Review
- Simplify or remove of SPI from doc
- Ask this to be WG work item.
Emergency-rsvp-00.txt
---------------------
Next Step
Janet Gunn:
The high number is the higher priority, this is reverse of
what is used in emergency.
Francois:
There are still typos in the draft, and this needs to be aligned with
other work to maintain consistency.
6) Matt Mathis 15
draft-ietf-tsvwg-tcp-mib-extension-08.txt
===================================================================
Matt presenting.
Did not present slides.
Have been through MIB doctors.
Hum indicates that when Matt is ready, start WGLC on this.
Meeting ended at 7:55 PM
|