Re: [mpls] Review of draft-ietf-mpls-lsp-ping-registries-update-01

["Carlos Pignataro (cpignata)" <[e-mail]>]

> 2020/04/02 午前7:13、Adrian Farrel <[e-mail]>のメール:
> 
> Thanks Loa,
> 
> I agree with your interpretation of 8126.
> 
> I think that the challenge with "experiments on the open Internet" is that the experiments have to have built into them some way to protect against two experiments using the same codepoint. That's not usually done in my experience, meaning that the two allocation classes are often pretty similar. Maybe there is some difference in duration of the use of a code point.
> 
> I'd certainly be happy with collapsing these registries to use just one range. I would say that keeping the resulting range small (just a few code points) is desirable.
> 

+1

Thanks,

Carlos.

> Best,
> Adrian
> 
> -----Original Message-----
> From: Loa Andersson <[e-mail]> 
> Sent: 02 April 2020 11:31
> To: [e-mail]; [e-mail]
> Cc: [e-mail]
> Subject: Re: Review of draft-ietf-mpls-lsp-ping-registries-update-01
> 
> Adrian,
> 
> This is to address your comment on "Private Use" and "Experimental Use",
> we will review the rest of the comments and update as needed.
> 
> On 02/04/2020 01:06, Adrian Farrel wrote:
>> Hi all,
>> 
> <snip>
>> 
>> I have a number of small editorials and some larger questions and issues
>> set out below. I also have one question that has broader scope:
>> 
>> For [IANA-MT] and [IANA-Sub-6] you now have both 'Private Use' and
>> 'Experimental Use'. I struggle to see how this makes sense. The uses
>> decribed in RFC 8126 are sufficiently similar that it is unusual to
>> have both categories defined for a single registry. I don't see anything
>> in the descriptive text in this document that makes clear why you need
>> both categories and how an implementation would decide which range to
>> select a code point from.
> <snip>
> 
> You are right I've been struggling with these two type of code points
> also, but came to a slightly different conclusion than you did.
> 
> RFC 8126 says:
> 
> 4.1.  Private Use
> 
>    Private Use is for private or local use only, with the type and
>    purpose defined by the local site.  No attempt is made to prevent
>    multiple sites from using the same value in different (and
>    incompatible) ways.  IANA does not record assignments from registries
>    or ranges with this policy (and therefore there is no need for IANA
>    to review them) and assignments are not generally useful for broad
>    interoperability.  It is the responsibility of the sites making use
>    of the Private Use range to ensure that no conflicts occur (within
>    the intended scope of use).
> 
>    Examples:
> 
>       Site-specific options in DHCP [RFC2939]
>       Fibre Channel Port Type Registry [RFC4044]
>       TLS ClientCertificateType Identifiers 224-255 [RFC5246]
> 
> 4.2.  Experimental Use
> 
>    Experimental Use is similar to Private Use, but with the purpose
>    being to facilitate experimentation.  See [RFC3692] for details.
>    IANA does not record assignments from registries or ranges with this
>    policy (and therefore there is no need for IANA to review them) and
>    assignments are not generally useful for broad interoperability.
>    Unless the registry explicitly allows it, it is not appropriate for
>    documents to select explicit values from registries or ranges with
>    this policy.  Specific experiments will select a value to use during
>    the experiment.
> 
>    When code points are set aside for Experimental Use, it's important
>    to make clear any expected restrictions on experimental scope.  For
>    example, say whether it's acceptable to run experiments using those
>    code points over the open Internet or whether such experiments should
>    be confined to more closed environments.  See [RFC6994] for an
>    example of such considerations.
> 
>    Example:
> 
>       Experimental Values in IPv4, IPv6, ICMPv4, ICMPv6, UDP, and TCP
>       Headers [RFC4727]
> 
> 
> It seems to me that "Private Use" are intended for private networks,
> where care is taken that the code points are not leaked into the
> Internet, but there the network itself is a production network, that
> will be run for an unforeseeable amount of time. And that "Experimental
> Use" code points are for short lived experiments.
> 
> 
> This is different.
> 
> I'm very uncertain whether it is sufficiently different to motivate two
> different types. If the working group thinks there should be only one
> code point, I would argue to keep the code points for "Experimental
> Use". If we converge on "one type of code point only, I think this has
> a wider impact than this document, and we should probably update RFC
> 8126 (again).
> 
> I'd like to invite comments on this on the mpls wg list.
> 
> /Loa
> 
> -- 
> 
> 
> Loa Andersson                        email: [e-mail]
> Senior MPLS Expert
> Bronze Dragon Consulting             phone: [phone]
>