Re: [sfc] MD Class registry

Re-,

Please see inline. 

Cheers,
Med

> -----Message d'origine-----
> De : sfc [mailto:sfc-bounces@ietf.org] De la part de Adrian Farrel
> Envoyé : lundi 20 février 2017 11:13
> À : 'Joel M. Halpern'; sfc@ietf.org
> Objet : Re: [sfc] MD Class registry - Vendor Extensibility
> 
> It seems to me, that we need to sort ourselves out a bit wrt flexibility
> in
> metadata.
> 
> We have a hierarchy:
> MD Class (2^16)
>    MD Type (2^7, about to become 2^8)
>       MD TLVs (different docs)
> 
> AFAICS only one "optional variable length metadata" can be present in an
> NSH.
> 
> So I have some observations...
> 
> o 2^16 is a lot of MD classes. Really a lot!

[Med] Agree.

> o 2^8 is more than plenty MD types in any class

[Med] At least for the mobile case where the practice to "enrich" packets with some data on-path (https://tools.ietf.org/html/draft-ietf-sfc-use-case-mobility-07#section-3.3) is the rule, the type of data that can be inserted may be very "rich"; see for example Section 7.1 of http://www.etsi.org/deliver/etsi_ts/129200_129299/129230/12.06.00_60/ts_129230v120600p.pdf. 

I know that is frightening but the experience from existing IETF protocols is that a 2^8 type registry can be exhausted: see for the example the RADIUS case http://www.iana.org/assignments/radius-types/radius-types.xhtml#radius-types-2.

In order to cover existing use cases while avoiding coming back to the IETF to claim the type resources have been exhausted, I do suggest to use more bits for the type field. 16 bits would be great.

> o A vendor-specific MD class only works on an SFP
>     made up *entirely* of products from that vendor
> o A vendor-specific MD type only works on an SFP
>     made up *entirely* of products from that vendor
> o The main (only) place we should care about vendor-
>     specific pieces of metadata is in the definitions of
>     TLVs in MC0/MD2 (not this doc).
> 
> Am I missing a particular use case?
> 
> Cheers,
> Adrian
> 
> > -----Original Message-----
> > From: sfc [mailto:sfc-bounces@ietf.org] On Behalf Of Joel M. Halpern
> > Sent: 20 February 2017 02:09
> > To: sfc@ietf.org
> > Subject: Re: [sfc] MD Class registry - Vendor Extensibility
> >
> > A related topic on the MD Class and MD Type Registry is that of Vendor
> > extensibility.  The current registry text reads:
> >
> >     0x0000 to 0x01ff: IETF Review
> >     0x0200 to 0xfff5: Expert Review
> >     0xfff6 to 0xfffe: Experimental
> >     0xffff: Reserved
> >
> > Even with the minor change I raised for discussion in my previous note,
> > the only space for Vendor's to define their own type values is in the
> > experimental space.  It seems to me that is not sufficient.
> >
> > How do we want to address this?  We could take a portion of the Class
> > space for vendors.  But how would we assign values to vendors for their
> use?
> >
> > Yours,
> > Joel
> >
> > _______________________________________________
> > sfc mailing list
> > sfc@ietf.org
> > https://www.ietf.org/mailman/listinfo/sfc
> 
> _______________________________________________
> sfc mailing list
> sfc@ietf.org
> https://www.ietf.org/mailman/listinfo/sfc