Michael Tuexen | 3 Jul 2007 10:46
Picon

Re: IANA issue with RFC2960/draft-ietf-tsvwg-2960bis

Hi Lars,

sounds good to me. Just two questions:
1. 2960bis is in the RFC editor queue. How will this change of text  
happen?
2. Do I need to write an ID or RFC to register a port? Or can I just  
contact
    IANA? I'm asking this only for registration of port numbers which  
are
    registered for TCP and/or UDP for a similar service like the discard
    port you use as an example.

Best regards
Michael

On Jun 28, 2007, at 7:53 PM, Lars Eggert wrote:

> Hi,
>
> during a second IANA review of the IANA considerations section in  
> draft-ietf-tsvwg-2960bis during the RFC Editor's "IANA" state, a  
> discrepancy has been identified. Basically, the text that describes  
> the port number assignment procedure - which had been taken  
> verbatim from RFC2960 - wasn't describing the procedure that IANA  
> has actually used since RFC2960 was published seven years ago.
>
> The IANA instructions in draft-ietf-tsvwg-2960bis and RFC2960  
> basically say "for each existing TCP/UDP port assignment, create a  
> similar entry in the SCTP port number registry", i.e., a blanket  
> assignment of SCTP port numbers for all existing TCP/UDP assignments.
(Continue reading)

Lars Eggert | 3 Jul 2007 10:53
Picon
Gravatar

Re: IANA issue with RFC2960/draft-ietf-tsvwg-2960bis

On 2007-7-3, at 11:46, ext Michael Tuexen wrote:
> sounds good to me. Just two questions:
> 1. 2960bis is in the RFC editor queue. How will this change of text  
> happen?

It's currently in the "IANA" state, and I asked IANA to hold it there  
until this issue is resolved. Once we have agreed on a procedure, we  
can update the text with an RFC Editor Note.

> 2. Do I need to write an ID or RFC to register a port? Or can I  
> just contact
>    IANA?

No ID needed, the idea is to use "First Come First Served" as an  
allocation policy, similar to UDP and TCP ports. I should probably  
make that explicit in the text.

(Also, note that the process IANA currently uses requires an Expert  
Review. I'm still trying to determine what the rationale for that was  
and if this should be continued. The current proposal does not have  
such an Expert Review stage.)

> I'm asking this only for registration of port numbers which are
>    registered for TCP and/or UDP for a similar service like the  
> discard
>    port you use as an example.

Under the current proposal, you don't ever need a draft, whether this  
is for a port number with an existing UDP/TCP registration or for one  
without.
(Continue reading)

Mark Allman | 3 Jul 2007 14:45

udp guidelines draft: small comments


Gorry, Lars-

I read through the UDP guidelines draft yesterday.  Overall I think it
is in good shape and to me it seems like it is getting close to ready to
forward.  I flagged some small things, which I list below.  No
showstoppers, IMO.

I hope this is helpful.

allman

Sec 1: "unreliable, message-passing" -> "unreliable, best-effort,
message-passing"

Sec 3: spell out "PMTU" on first use

Sec 3.1: The first paragraph is a bit fuzzy, I think.  In particular,
you are advocating that all traffic to a "destination" be controlled
together.  This brings to mind two questions.  First, what is a
"destination"?  A host?  A host+port?  Second, this isn't how the
nominal congestion controller (TCP) does things.  I.e., you say that if
multiple processes are forked then CC should be performed across all of
them (assuming their traffic is going to the same "destination").  That
seems to be different from what TCP might do in this case.  I would like
to see this guidance be more crisp and concrete.  (That said, I am
sitting here at a sort of loss for how to do it.  Maybe I just need more
coffee.)

Sec 3.1.1: "SHOULD consider implementing TCP-friendly" -> "SHOULD
(Continue reading)

Lars Eggert | 3 Jul 2007 15:09
Picon
Gravatar

Re: udp guidelines draft: small comments

Hi,

thanks for the feedback. I'll fix the nits you have identified, but  
there is one issue that the WG should discuss some more:

On 2007-7-3, at 15:45, ext Mark Allman wrote:
> Sec 3.1: The first paragraph is a bit fuzzy, I think.  In particular,
> you are advocating that all traffic to a "destination" be controlled
> together.  This brings to mind two questions.  First, what is a
> "destination"?  A host?  A host+port?  Second, this isn't how the
> nominal congestion controller (TCP) does things.  I.e., you say  
> that if
> multiple processes are forked then CC should be performed across  
> all of
> them (assuming their traffic is going to the same "destination").   
> That
> seems to be different from what TCP might do in this case.  I would  
> like
> to see this guidance be more crisp and concrete.  (That said, I am
> sitting here at a sort of loss for how to do it.  Maybe I just need  
> more
> coffee.)

The motivation for this text came from seeing a number of request- 
response protocols, where a client shoots off many parallel requests  
to the same server, maybe from different threads or processes. UDP  
has no connections and so it's easy to say "well, it's just one  
request and one response per (application-level) connection, so  
there's no way to do CC", which is inaccurate. Some of these  
protocols initially talked about allowing hundreds of parallel  
(Continue reading)

Michael Tuexen | 3 Jul 2007 16:02
Picon

Re: IANA issue with RFC2960/draft-ietf-tsvwg-2960bis

Hi Lars,

then I'm fine with your proposal.

Best regards
Michael

On Jul 3, 2007, at 10:53 AM, Lars Eggert wrote:

> On 2007-7-3, at 11:46, ext Michael Tuexen wrote:
>> sounds good to me. Just two questions:
>> 1. 2960bis is in the RFC editor queue. How will this change of  
>> text happen?
>
> It's currently in the "IANA" state, and I asked IANA to hold it  
> there until this issue is resolved. Once we have agreed on a  
> procedure, we can update the text with an RFC Editor Note.
>
>> 2. Do I need to write an ID or RFC to register a port? Or can I  
>> just contact
>>    IANA?
>
> No ID needed, the idea is to use "First Come First Served" as an  
> allocation policy, similar to UDP and TCP ports. I should probably  
> make that explicit in the text.
>
> (Also, note that the process IANA currently uses requires an Expert  
> Review. I'm still trying to determine what the rationale for that  
> was and if this should be continued. The current proposal does not  
> have such an Expert Review stage.)
(Continue reading)

Bruce Davie | 3 Jul 2007 20:12
Picon
Favicon

draft-davie-tsvwg-rsvp-l3vpn-00.txt

I've posted a draft that may be of interest to the TSVWG and L3VPN  
working groups. While it percolates through the ID system, you can  
get a copy from

ftp://ftpeng.cisco.com/bdavie/draft-davie-tsvwg-rsvp-l3vpn-00.txt

I've asked for a slot in both WGs in Chicago, with the intention of  
deciding afterwards which group (if any) might be the more  
appropriate home.

Abstract

    RFC 4364 defines an approach to building provider-provisioned  
Layer 3
    VPNs.  It may be desirable to use RSVP to perform admission control
    on the links between CE and PE routers.  This document specifies
    procedures by which RSVP messages travelling from CE to CE across an
    L3VPN may be appropriately handled by PE routers so that admission
    control can be performed on PE-CE links.  Optionally, admission
    control across the provider's backbone may also be supported.

Lars Eggert | 4 Jul 2007 12:34
Picon
Gravatar

Re: Re: [tcpm] Revision of draft-larsen-tsvwg-port-randomization

On 2007-5-31, at 17:51, ext Lars Eggert wrote:
> The concepts in this draft are likely relevant to most of our  
> transport protocols, and hence would be in scope for TSVWG. The  
> TSVWG chairs are interested in comments on whether there is group  
> interest in this draft - please comment on tsvwg <at> ietf.org.

We've received some positive feedback on adopting this draft, but I'd  
like to see a stronger show of support, because this draft impacts  
several of our transport protocols at the same time.

Please comment on tsvwg <at> ietf.org - reply-to set accordingly.

Lars

arif khan | 4 Jul 2007 17:31
Picon

Configuration of 2 local as with single asp in signle exchange

Hi,
 
i want to rout message using two opc=X and opc=Y using m3ua stack configuration of adding two local AS with opc X and opc Y to dpc Z.
 
The side having opc X and Y and the side having dpc Z is in ASP-ASP configuration in single exchange both the side.
I added following configuration at m3ua:
side X and Y:
==========
added asp (aspid 1)
added remote asp(remaspid 1)
added remote as (remote asid=1,rc=1,dpc=Z)
added local as (asid= 1,rc=1,opc=X)
added local as (asid=2,rc=2,opc=Y)
 
side z:
=======
added asp (aspid 1)
added remote asp(remoteaspid=1)
added remote as (remoteasid=1,rc=1,dpc=X)
added remote as (remoteasid=2,rc=2,dpc=Y)
added local as (asid=1,rc=1,opc=Z)
 
what i am missing in the configuration and what should be the correct configuration according to rfc and maintaining interoperability of the asp-asp in single exchange.
 
tia
ar
 
 
Francois Le Faucheur IMAP | 5 Jul 2007 15:02
Picon
Favicon

Re: RSVP proxy approaches

Chris, Fred, Pratik and all,

Please see below:

On 19 Jun 2007, at 20:40, Christou, Christos wrote:

Francois,
 
Yes, I do think this works.
 
Thanks,
 
Chris

From: Francois Le Faucheur IMAP [mailto:flefauch <at> cisco.com]
Sent: Tuesday, June 19, 2007 11:22 AM
To: Christou, Christos
Cc: Francois Le Faucheur IMAP; tsvwg <at> ietf.org
Subject: Re: [Tsvwg] RSVP proxy approaches

Chris and all,

I'm started editing rsvp-proxy-approaches to reflect all discussed changes.

Regarding the input below, I would propose that we incorporate it as a new section under Appendix A (eg "A.5 RSVP-based Voice/Video CAC in IPsec environments"). This seems reasonable since:
* Appendix A is dedicated to describing use cases for the RSVP Proxy approaches described in the main sections of the I-D .
* the fundamental approach used in this scenario is already described in the main sections of the I-D. The text below focuses on some details specific to when the approach is used in the presence of IPsec. 


Below is the text I put together for the new Appendix A.5.
I added material to explain how this relates to [I-D.ietf-tsvwg-vpn-signaled-preemption].
If you want to send comments/suggestions by the end of today I could address those before posting.
Francois



A.5.  RSVP Proxies for Reservations in the presence of IPsec Gateways

   [I-D.ietf-tsvwg-vpn-signaled-preemption] discusses how resource
   reservation can be supported end-to-end in a nested VPN environment.
   At each VPN level, VPN Routers behave as [RFC4301] security gateways
   between a plaintext domain and a cyphertext domain.  To achieve end-
   to-end resource reservation, the VPN Routers process RSVP signaling
   on the plaintext side, perform aggregation of plaintext reservations,
   and maintain the corresponding aggregate RSVP reservations on the
   cyphertext side.  Each aggregate reservation is established on behalf
   of multiple encrypted end-to-end sessions sharing the same ingress
   and egress VPN Routers.  These aggregate reservations can be as
   specified in [RFC3175] or [RFC4860].

   Section 3 of [I-D.ietf-tsvwg-vpn-signaled-preemption] discusses the
   necessary data flows within a VPN Router to achieve the behavior
   described in the previous paragraph.  Two mechanisms are described to
   achieve such data flows.  Section 3.1 presents the case where the VPN
   Router carries data across the cryptographic boundary.  Section 3.2
   discusses the case where the VPN router uses a Network-Guard.

   Where such mechanisms are not supported by the VPN Routers, the
   approach for end-to-end reservation presented in
   [I-D.ietf-tsvwg-vpn-signaled-preemption] cannot be deployed.  An
   alternative approach to support resource reservations within the
   cyphertext core is to use the "Application_Entity-Controlled Proxy"
   approach (as defined in Section 4.5) in the following way:

   o  the RSVP Proxies are located inside the cyphertext domain and use
      aggregate RSVP reservations,

   o  the Application Entity exchange application level signaling with
      the end systems in the plaintext domain,

   o  the Application Entity controls the RSVP Proxies in the cyphertext
      domain via an RSVP Proxy control interface

   This is illustrated in Figure 17 in the case where the application is
   SIP-based multimedia communications.



         |-------|                                    |-------|
         |SIP    |///////////////////\\\\\\\\\\\\\\\\\|SIP    |
        /|Server/|                                    |Server/|\
       / |Proxy  |                                    |Proxy  | \
      /  |-------|                                    |-------|  \
     /      ^    \\                                  //   ^       \
    /       ^     \\                                //    ^        \
   /        ^      \\                              //     ^         \
|---|   |------|  |--------|   ***   ***   |--------|  |------|   |---|
| S |---|IPsec |--|  ARSVP |---*r*---*r*---| ARSVP  |--|IPsec |---| R |
|---|   | GW   |  | Sender |   ***   ***   |Receiver|  | GW   |   |---|
        |------|  |  Proxy |               | Proxy  |  |------|
                  |--------|               |--------|

     ***PT*****> **********************CT****************> ****PT***>

     =====>                                                   =====>

                            =====ARSVP======>



|----| RSVP-capable      |----| RSVP-capable         ***
| S  | Sender            | R  | Receiver             *r* regular RSVP
|----|                   |----|                      *** router

|------|
|IPsec | IPsec gateway
| GW   |
|------|

ARSVP Aggregate RSVP

***>  media flow

==>   segment of flow path protected by RSVP reservation

/ \   SIP signaling

  ^    Network management interface between SIP Server/Proxy
       and IPsec device

//    control interface between SIP Server/Proxy and ARSVP Proxy

PT    Plaintext network

CT    Cyphertext network


     Figure 17: RSVP Proxies for Reservations in the Presence of IPsec
                                 Gateways

   Where the sender and receiver are RSVP capable, they may also use
   RSVP signaling.  This achieves resource reservation on the plaintext
   segments of the end-to-end i.e. :

   o  from the sender to the ingress IPsec gateway and

   o  from the egress IPsec gateway to the receiver.

   In this use case, because the VPN Routers do not support any RSVP
   specific mechanism, the end-to-end RSVP signaling is effectively
   hidden by the IPsec gateways on the cyphertext segment of the end-to-
   end path.

   As with the "Application_Entity-Controlled Proxy" approach (defined
   in Section 4.5), the solution here for synchronizing RSVP signaling
   with application-level signaling is to rely on an application-level
   signaling device that controls an on-path RSVP Proxy function.
   However, in the present use case, the RSVP Proxies are a component of
   a cyphertext network where all user (bearer) traffic is IPsec
   encrypted.  This has a number of implications including the
   following:

   1.  encrypted flows can not be identified in the cyphertext domain so
       that network nodes can only classify traffic based on IP address
       and DiffServ Code Points (DSCPs).  As a result, only aggregate
       RSVP reservations (such as those specified in [RFC3175] or
       [RFC4860] ) can be used.  This is similar to
       [I-D.ietf-tsvwg-vpn-signaled-preemption].

   2.  Determining the RSVP Sender proxy and RSVP receiver Proxy to be
       used for aggregation of a given flow from sender to receiver
       creates a number of challenges.  Details on how this may be
       achieved are beyond the scope of this document.  We observe that,
       as illustrated in Figure 17, this may be facilitated by a network
       management interface between the application entity and the IPsec
       gateways.  For example, this interface may be used by the
       application entity to obtain information about which IPsec
       gateway is on the path of a given end-to-end flow.  Then, the
       application entity may maintain awareness of which RSVP Proxy is
       on the cyphertext path between a given pair of IPsec gateways.
       How such awareness is achieved is beyond the scope of this
       document.  We simply observe that such awareness can be easily
       achieved through simple configuration in the particular case
       where a single (physical or logical) Proxy device is fronting a
       given IPsec gateway.  We also observe that when awareness of the
       RSVP Receiver Proxy for a particular egress IPsec gateway (or
       end-to-end flow) is not available, the aggregate reservation may
       be signaled by the RSVP Sender Proxy to the destination address
       of the egress IPsec gateway and then proxied by the RSVP Receiver
       Proxy.

   Different flavors of operations are possible in terms of aggregate
   reservation sizing.  For example, the application entity can initiate
   an aggregate reservation of fixed size a priori and then simply keep
   count of the bandwidth used by sessions and reject sessions that
   would result in excess usage of an aggregate reservation.  The
   application entity could also re-size the aggregate reservations on a
   session by session basis.  Alternatively, the application entity
   could re-size the aggregate reservations in step increments typically
   corresponding to the bandwidth requirement of multiple sessions.




Julian Satran | 5 Jul 2007 19:05
Picon
Favicon

Julian Satran/Haifa/IBM is out of the office.


I will be out of the office starting  05/07/2007 and will not return until
15/07/2007.

I have no access to e-mail. If your message is important  please resend it
close to July 17. For urgent matters you may contact me through my
assistant - Daphne Maor-Belsky.


Gmane