Re: [rbridge] draft-ietf-trill-rbridge-oam-02

[somnath chatterjee <[e-mail]>]

Hi David,

Please ignore my 3rd point regarding sequence numbers, I missed the
random/arbitrary generation of sequence numbers in the draft. With
random generation of 28bits sequence number the probability of
collision tends to zero.

Thank you,
Regards,
Somnath

On Wed, Mar 21, 2012 at 7:11 PM, David M Bond <[e-mail]> wrote:
> Hello Somnath,
>
> Thanks for the comments. My responses are inline below with [DBOND].
>
> Thanks,
> David
>
>
>
> From:   somnath chatterjee <[e-mail]>
> To:     [e-mail]
> Date:   03/21/2012 07:58 AM
> Subject:        [rbridge]  draft-ietf-trill-rbridge-oam-02
> Sent by:        [e-mail]
>
>
>
> Hi Authors,
>
> 1>In Section 4.1.1 two  statements can be found which is ambiguous.
> In first paragraph
> "The M bit MUST be zero."
> and in sixth paragraph
> "The M bit MUST be zero for a known unicast ping."
>
> I think the ping tool should not be limited to unicast use only and
> should be allowed to be sent on distribution tree when used for
> multidestination frames/unknown unicast. It would be good to include
> excerpts from draft-yizhou-trill-multi-destination-ping in this OAM
> draft which talks about multi-destination ping, jitters etc.
>
> [DBOND] Agreed. In fact I have been talking with Yizhou about merging these
> two drafts but I have not had a chance yet. One item that did come up that
> is useful to mention is in relation to jitter. Currently in Yizhou's draft
> she has an alternative format for echos so that the time stamp is in a
> fixed offset for hardware. If at all possible I would prefer to have a
> single format for echos. My idea was to say time stamps are TLVs but if
> they appear they MUST always be the first TLV in the TLV list. This way
> they have a fixed location in the packet so hardware could work with them.
> Could some hardware folks comment on the viability of this? In general I
> see there being three options here:
>
>  Alternative Format Echo with TimeStamp
>  Change Base Format of Echo to include TimeStamp
>  Fixed location of TimeStamp TLV
>
> What does everyone in the WG think?
> [/DBOND]
>
> 2> It would be good if RBridges that implement TRILL- OAM does not
> decrement Hop count by more than one for OAM/Channel messages only.
> This would help us limiting the max hop count used for traceroute to
> be equal to the diameter of the campus.
>
> [DBOND] To do this there would have to be separate logic for TRILL OAM
> messages. It is very important that OAM frames follow the same logic as any
> other frame (to test the forwarding plane) and so we can't make a special
> condition as such.[/DBOND]
>
> 3> If my understanding is correct there is no way to distinguish
> between pings reply and unicast traceroute reply(other than the error
> replies). An user may use both the tool concurrently or with minimal
> time gap. As such it would be hard to distinguish between the replies
> only with sequence number. I would suggest to use some unique
> identifier to differentiate between the two types of messages.
>
> [DBOND] I'm unsure why the sequence number would be insufficient to meet
> this need. Please elaborate.
>
> 4> In Section 4.1.2.1
> ". . . various forms of ECMP hashing based on fields such as MAC
> addresses, IP addresses, and/or TCP/UDP port numbers."
>  Vlans may be added to the fields list.
>
> [DBOND] Adding
>
> 5> In Section 4.1.2.2
> Editorial changes for "One cannot use the diemater of the network . .
> ."  should be replaced by
> "One cannot use the diameter of the network/campus"
>
> [DBOND] Adding
>
> Thank you,
> Regards,
> Somnath
> _______________________________________________
> rbridge mailing list
> [e-mail]
> http://mailman.postel.org/mailman/listinfo/rbridge
>
>
>

_______________________________________________
rbridge mailing list
[e-mail]
http://mailman.postel.org/mailman/listinfo/rbridge