On Tue, May 16, 2017 at 10:29 PM, Brian E Carpenter
<[email protected]> wrote:
> On 14/05/2017 05:42, Tom Herbert wrote:
>> Hello,
>>
>> At the Chicago WG meeting I made a request that ILA be taken up as a
>> WG item in int-area. The WG chairs and AD requested that we raise a
>> discussion on the list about what else is needed to be done for ILA
>> (Identifier Locator Addressing draft-herbert-nvo3-ila-04). The
>> question was also raised if int-area is the right WG for ILA or if it
>> should have a BOF.
>
> I think it's fine for int-area to take this up, as long as 6man
> is alerted.
>
> A few somewhat random comments:
>
> Is there any particular reason you chose a /64 routing prefix,
> rather than say /72 or /80? I don't see any dependency on SLAAC,
> so there's no reason to care about /64, right? I assume you would
> use DHCPv6 or some other method of assigning addresses to VMs.
>
The 64 bit was for convenience since that is what ILNP had and in the
DC we wanted enough identifier bits to be able to do identifier
assignment based on a timestamp as described in appendix B. There's no
hard requirement for it. I think we could just say an ILA domain can
define lengths of identifier and locators as they see fit-- as long as
its consistent and unambiguous within the domain there shouldn't be an
issue.

> If you used /72 or /80 you'd have flexibility with a ULA prefix to
> insert a subnet field.
>
>> The format of an ILA address with a global unicast locator is:
>>
>>    |<--------------- Locator --------------->|
>>    |3 bits| N bits        | M bits  | 61-N-M | 64 bits              |
>>    +------+-------------+---------+---------------------------------+
>>    | 001  | Global prefix | Subnet  | Host   |      Identifier      |
>>    +------+---------------+---------+--------+----------------------+
>
> Why is this restricted to 2000::/3?

Probably shouldn't be. If locator length isn't fixed like above, then
probably should just describe it by properties of being routable to
physical hosts.

>
>> The format of an ILA address with a unique local IPv6 unicast locator
>> is:
>>
>>    |<--------------- Locator --------->|
>>    | 7 bits |1|  40 bits   |  16 bits  |          64 bits           |
>>    +--------+-+------------+-----------+----------------------------+
>>    | FC00   |L| Global ID  | Host      |        Identifier          |
>>    +--------+-+------------+-----------+----------------------------+
>
> The first byte is confusing; presumably you're trying to express
> fd00::/8 so why not put 11111101 ?
>
> I agree that the flow label would be tricky to use for this sort of
> purpose, but I have a couple of comments anyway:
>
>>   o The flow label is only twenty bits, this is too small to be a
>>     discriminator in forwarding a virtual packet to a specific
>>     destination. Conceptually, the flow label might be used in a
>>     type of label switching to solve that.
>
> Perhaps, but 6man fairly comprehensively rejected that type of usage.
> In practice it's excluded by section 4 of RFC6437.
>
>       o The flow label is not considered immutable in transit,
>         intermediate devices may change it.
>
> That's not phrased quite right. RFC6437 section 2 says:
>
> "  Once set to a non-zero value, the Flow Label is expected to be
>    delivered unchanged to the destination node(s).  A forwarding node
>    MUST either leave a non-zero flow label value unchanged or change it
>    only for compelling operational security reasons as described in
>    Section 6.1.
>
>    There is no way to verify whether a flow label has been modified en
>    route..."
>
> You could in fact have an operational guarantee that within an
> administrative domain, no middleboxes will change the flow label.
> So something like this is more accurate:
>
>      o  The flow label is not protected against changes in transit.
>
Yes, the flow label concept came up in some early dicussions.

>>    o Intermediate devices must not insert extension headers
>>      [RFC2460bis].
>
> I think you could say:
>
>       o Current specifications do not allow intermediate devices to
>         insert extension headers [RFC2460bis].
>
Good point.

> Regards
>    Brian

_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area

Reply via email to