Re: [mpls] "Traffic Management" (was Re: Poll on renaming of EXP field)

["Eric Gray" <[e-mail]>]

George,

	In response to your comment - with chair hat on - the
fact that there is no registry has not affected the way we
have proceeded (within the IETF) so far.  Every use of the
EXP bits that we currently consider "legitimate" is defined
in an RFC.

--
Eric Gray
Principal Engineer
Ericsson  

> -----Original Message-----
> From: George Swallow [mailto:[e-mail]] 
> Sent: Tuesday, August 19, 2008 3:13 PM
> To: Eric Gray; Francois Le Faucheur IMAP
> Cc: [e-mail]
> Subject: Re: [mpls] "Traffic Management" (was Re: Poll on 
> renaming of EXP field)
> Importance: High
> 
> WG Chair hat on:
> 
> I think it is important to signal the change by changing the 
> name.  Just
> clarifying the meaning of Experimental may result in a 
> document that does
> not get read.  Further, there is no registry for the EXP bit 
> usage and that
> usage can vary between LSPs running within the same network.  
> So comparing
> this to anything that has a registry and a process of IETF 
> consensus does
> not apply well here.
> 
> I'm also believe that when this is published as an RFC it 
> will say that it
> updates the appropriate RFCs.  Loa, please correct me if I am wrong.
> 
> WG Chair hat off:
> 
> One of the problems here is that every term used in this 
> general space is
> already over-loaded.  Further the emphasis implied by a 
> particular term
> varies with context (and consequently with the reader's 
> background).  For
> instance the term Traffic Management has very strong 
> connotations for those
> involved in the ATM Forum.
> 
> My personal view is that the term CoS is good enough.  But I 
> do believe that
> type of service (ToS) would have been an improvement.  It's 
> not that I think
> "type" is a better term than "class", it is that this would 
> make it the same
> as the IPv4 header.  The ToS field in the IP header now 
> encompasses both the
> DS codepoint and the ECN codepoint.  These are precisely the 
> functions that
> we are now trying to squeeze into three bits of the MPLS 
> Label Stack Entry.
> 
> I would argue strongly for this change were it not that in 
> IPv6 the same
> function is carried in a field called Traffic Class.
> 
> So CoS takes a little from each of v4 and v6.  Good enough.
> 
> ...George
> 
> 
> On 8/19/08 1:53 PM, "Eric Gray" <[e-mail]> wrote:
> 
> > Francois,
> > 
> > If I didn't care about the potential for extra work (and
> > possibly confusion as well), I would think "Traffic Management"
> > a better (because more generic) field name than "CoS" - but it
> > is still a change that may have far reaching consequences, and
> > which is better handled by a better explanation of the meaning
> > of the field than by any effort to come up with a better name
> > for it.
> > 
> > However, even the field-name "Traffic Management" may
> > not be generic enough.
> > 
> > At present, we seem to be agreed that currently defined
> > legitimate uses for the field are all related (at least in a
> > stretch) to something to do with how frames are expected to be
> > handled in forwarding.  But - given a precedent established in
> > RFC 3270 (which, among other things, emphasizes that meaning or
> > semantics of the field depends on a common understanding of why
> > an LSP was established) - it is actually clear that the field
> > might mean other things as well.  For example, the field might
> > - in some future context - be used to indicate error handling
> > for packets that will be dropped, or macro statistics buckets
> > that individual packets belong to (unrelated to other handling
> > of a packet so marked).
> > 
> > In fact, we're re-hashing many of the same arguments that
> > led to the field's being named as it was.
> > 
> > So, the issue is all about what constitutes a "legitimate
> > use" - which I would argue is defined by IETF consensus, and is
> > an issue better handled by good explanation than by renaming.
> > 
> > --
> > Eric Gray
> > Principal Engineer
> > Ericsson  
> > 
> >> -----Original Message-----
> >> From: [e-mail] [mailto:[e-mail]] On
> >> Behalf Of Francois Le Faucheur IMAP
> >> Sent: Tuesday, August 19, 2008 12:42 PM
> >> To: George Swallow
> >> Cc: [e-mail]; Francois Le Faucheur IMAP
> >> Subject: [mpls] "Traffic Management" (was Re: Poll on
> >> renaming of EXP field)
> >> 
> >> Hi,
> >> 
> >> Someone earlier suggested "Traffic Management". While it can most
> >> certainly be argued this would not be a perfect name either
> >> (or other  
> >> names would be closer to "perfection"), is there a good
> >> argument I am  
> >> missing for why "COS" is obviously better (or more "good enough")
> >> than "Traffic Management"? (considering we want to capture both
> >> Diffserv and ECN/PCN use of EXP field)
> >> 
> >> Thanks
> >> 
> >> Francois
> >> 
> >> 
> >> On 18 Aug 2008, at 22:23, George Swallow wrote:
> >> 
> >>> During the last call on "EXP field" renamed to  "CoS Field"
> >>> draft-ietf-mpls-cosfield-def-04.txt, there were comments on
> >>> alternatives to the name COS.
> >>> 
> >>> This message initiates a two week poll on whether the name COS
> >>> is good enough, or if some other name is needed.  The poll closes
> >>> 23:59 Sept 1 GMT.
> >>> 
> >>> Please answer with a simple yes or no.  You may send any
> >> additional  
> >>> comment
> >>> in a separate message (with a different subject line).
> >>> 
> >>> ...George
> >>> 
> >>> _______________________________________________
> >>> mpls mailing list
> >>> [e-mail]
> >>> https://www.ietf.org/mailman/listinfo/mpls
> >> 
> >> _______________________________________________
> >> mpls mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/mpls
> >> 
> 
> 
_______________________________________________
mpls mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/mpls