Michael,

I get a bit confused by the way your mail agent doesn't distinguish
new and old text, but in line...

On 24/09/2017 08:35, Michael Richardson wrote:
> 
> Toerless, thank you for this version.  We are very very close!
> Brian, Max, I invoke your name in my comments... please read and +1,
> or disagree with me.
> 
> I noticed in this version has two places with no space after the .:
> 
>          across the network.[RFC7575] defines the fundamental ideas and design
>            and it should be re-usable by all autonomic functions.[RFC7575]
>            calls
> 
> 
> Can we avoid semantic/normative statments like "ACP routing table may include
> multiple" in a terminology section?
> 
> Change:
> ACP (ULA) prefix(es)  The prefixes routed across the ACP.  In the
>             normal/simple case, the ACP has one ULA prefix, see Section 6.10.
>             The ACP routing table may include multiple ULA prefixes if the
>             "rsub" option is used to create addresses from more than one ULA
>             prefix.  See Section 6.1.1.  The ACP may also include non-ULA
>             prefixes if those are configured on ACP connect interfaces.  See
>             Section 8.1.1.
> 
> to:
>   ACP (ULA) prefix(es)  The prefixes routed across the ACP.  In the
>             normative case, the ACP has one ULA prefix as specified in
>               Section 6.10.  The cases where multiple prefixes are present
>               are explained in Section 6.1.1, and use of non-ULA prefixes
>               are explained in Section 8.1.1.
> 
> There are some good updates to the ACP domain.  I want to bring up this issue
> From BRSKI here:
>            https://github.com/anima-wg/anima-bootstrap/issues/20
> 
> I was suggesting that the information for the Pledge's CSR should be broken
> up such that the pledge will create the right CN from the pieces, which it
> needs to know anyway. I had suggested:
> 
>    "ACPinfo" : { "acp-prefix" : "fda379A6f6ee00000200000064000001",
>                     "acp-prefix-length": 96,
>                     "acp-plus-part": "area51.research",
>                     "acp-domain": "acp.example.com"
>    }
> 
> I don't like your text:
>            If the operator does not own any FQDN, it should
>          choose am FQDN format string that intends to be equally unique.
> 
> I suggest that:
>   If they don't own any FQDN, then, aside from WTF?, that they form a
>   unique name as: $(uuidgen).example.com.
>   Or, perhaps that they use ${ULA}.example.com.
> 
> the only real-life situation where this is going to happen is in a test lab
> run by co-op students, and in that case, the software might as well have a
> pre-configured default.
> 
> ======
> 
> Brian, Max and I asked you to remove the EST requirements on announcements
> From the ACP document.

Toerless explained elsewhere why he thinks the duplication is needed.
 
> ======
>    ACP secure channel MUST imediately be terminated when the lifetime of
>          any certificate in the chain used to authenticate the neighbor
>          expires or becomes revoked.  Note that is is not standard behavior in
>          secure channel protocols such as IPsec because the certificate
>          authentication only influences the setup of the secure channel in
>          these protocols.
> 
> This is rather impractical to implement in real life, that's why IPsec
> doesn't do that.  I strongly suggest you sit down with some Cisco, Juniper
> (Netscreen) and Checkpoint IPsec product managers, and ask them what they
> actually do, and why.
> 
> Assuming you really want to have this kind of behaviour, then the correct way
> to do this is to make statements about the rekey intervals on the channels.
> 
>    ACP secure channels created with the use of certificates will include some
>        lifetimes for the certificates. The set of lifetimes includes:
>                  1) the expiry date of the certificate
>                  2) the lifetime of the OCSP response
>                  3) the validity time of the CRL response
> 
>        The earliest date provided by these provides a time at which the
>        keymanagement channel (e.g. IKEv2 PARENT SA) SHOULD be rekeyed.
> 
> 
> 6.10.4.  ACP Manual Addressing Sub-Scheme
> 
>    The sub-scheme defined here is defined by the Type value 1 (one) in
> 
> was changed to:
>    The sub-scheme defined here is defined by the Type value 00b (zero)
> 
> I think that this is a typo, and should be 01b ?
> Or does the Manual mechanism somehow fit into the
>    ACP Zone Addressing Sub-Scheme because the Z bit is different?
> 
> Do you explain this manual zone better somewhere?
> 
> 
> Can you show the two Types with their bits set at the end of the base scheme?
> 
>                 64                             64
>     +---------------------+---------+---++-----------------------------+
>     |    (base scheme)  01|Subnet-ID| Z ||     Interface Identifier    |
>     +---------------------+---------+---++-----------------------------+
>     <-------48--------->TT    13      1
> 
> 
> section 8.8.1, starting at:
>         The ACP connect interface must be (auto-)configured with an IPv6
> 
> seems to be missing articles and/or words, and needs an english editing pass.
> 
> ====== section 11
>            This document may be considered to be updating the IPv6 addressing
>          architecture ([RFC4291]) and/or the Unique Local IPv6 Unicast
>          addresses ([RFC4193]) depending on how strict specific statements in
> 
> I don't like this statement.  Either it violates the spec, and updates it, or
> it does not.  I do not think that it violates.  This is not a multi-link
> subnet, this is a prefix that is not on-link.

As readers of the 6man list know, this has been a very contentious topic.
I think it's safer to duck it in the ACP draft: say what we do, but say
nothing about RFC4291 etc.

>         It is possible, that this scheme constitutes an update to RFC4191
>       because the same 64 bit subnet prefix is used across many ACP
>       devices.  The ACP Zone addressing Sub-Scheme is very similar to the
>       common operational practices of assigning /128 loopback addresses to
>       network devices from the same /48 or /64 subnet prefix.
> 
> It does not. Brian? Do you concur?

Put it this way. The ACP doesn't assign ULAs using DHCPv6. It doesn't assign
them using SLAAC. It doesn't use conventionally sized /64 subnets. So in that
sense it is like RFC 6164 ("Using 127-Bit IPv6 Prefixes on Inter-Router Links").
But we don't need to apologise. As you say, we're simply assigning /128
addresses to (virtual) interfaces using our own scheme. And we're relying
on BCP 198 (RFC 7608) which says that routing prefixes can be any length
up to /128. If you want to cite anything, cite RFC 7608.

>         The goal for the 8 or 16-bit addresses available to an ACP device in
>         this scheme is to assign them as required to software components,
>         which in autonomic networking are called ASA (Autonomic Service
> 
> We are not providing 8-bit or 16-bit IIDs.
> We are providing 256 or 65536 /128 addresses which are conveniently
> aggregated for routing purposes.
> 
>          In practical terms, the ACP should be enabled to create from its /8
>          or /16 prefix one or more device internal virtual subnets and to
>          start software components connected to those virtual subnets.
> 
> No, don't say this, and don't do this in practice.  Create /128 routes to LL
> address of the internal VM and configure the /128 as a loopback address
> inside the VM.

So yes, I concur with Michael.

    Brian

_______________________________________________
Anima mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/anima

Reply via email to