RE: [dhcwg] dhcpv6-24: Interface-ID option

["Bernie Volz (EUD)" <[e-mail]>]

Thomas:

The link-address field is always used by the server to determine what link the client is on.

The interface-id is for use by the relay to help it return the response from the server to the client.

The reasons for having this include:
- It could be that the link-address field isn't specific enough. Perhaps there are a bunch of circuits all with one prefix (the link-address) but the response must be sent to the client over the correct circuit. Hence, the circuit information can be stored by the relay in the interface-id option.
- Even if the link-address is sufficient, the interface-id could be used as an optimization by the relay.

So:

>If it is
>included, but there is no valid link-address (which I understand is
>the reason for defining this particular option), how does the server
>know which addresses to assign the client? I.e., how does the server
>know on which link the client is?

No, there must always be a valid link-address. It just isn't specific enough.

>    The relay agent puts the client's address in the link-address field
>    regardless of whether the relay agent includes an Interface-id option
>    in the Relay-forward message.

Agreed. This is wrong and needs to be fixed. It should say:

   The relay agent puts a global or site-scoped address with a prefix
   assigned to the link on which the client should be assigned an
   address in the link-address field regardless of whether the relay
   agent includes an Interface-id option in the Relay-forward message.

(This is essentially the text in 20.1).

>Shouldn't the first MAY be a MUST? I.e., the link-address field
>contains no useful information.

No. The server only needs to the link-address to do address assignment in general. There could be instances where a specific relay/server arrangement exists and the server might want to consider the Interface-ID value (for classing, access control, etc).

>Any reaosn not to make this 32 bits? the IPv6 API already has 32-bit
>interface IDs. Is there any reason to make it larger?

I think it best to keep it variable-length since relays may use this for optimization reasons. For example, in 3G specs we were thinking of allowing the PDP-Context number on the GGSN to be carried in this field (this speeds the return of the reply to the client). While this might fit into 32-bits, why force it?

- Bernie

-----Original Message-----
From: Thomas Narten [mailto:[e-mail]]
Sent: Wednesday, May 08, 2002 11:10 AM
To: [e-mail]
Subject: [dhcwg] dhcpv6-24: Interface-ID option


>    If the relay agent cannot use the address in the link-address field
>    to identify the interface through which the response to the client
>    will be forwarded, the relay agent MUST include an Interface-id
>    option (see section 22.19) in the Relay-forward message.  The server

I think the interface-id option may be underspecified. If it is
included, but there is no valid link-address (which I understand is
the reason for defining this particular option), how does the server
know which addresses to assign the client? I.e., how does the server
know on which link the client is?

>    The relay agent puts the client's address in the link-address field
>    regardless of whether the relay agent includes an Interface-id option
>    in the Relay-forward message.

This doesn't seem right to me. I thought that the link-address is
supposed to hold the address of the relay agent. Seems wrong to put
the client's address in it. The link-address field is what the server
uses to figure out which link the client is on (right?). Also, since
the client's address is a link-local address, this field doesn't seem
to contain useful information for the server in this case.

>    Servers MAY use the Interface-ID for parameter assignment policies.
>    The Interface-ID SHOULD be considered an opaque value, with policies
>    based on exact string match only; that is, the Interface-ID SHOULD
>    NOT be internally parsed by the server.

Shouldn't the first MAY be a MUST? I.e., the link-address field
contains no useful information.

Also, not sure the above is sufficient. Unless the
interface-identifier is somehow stable, it's not clear how the server
policy could take it into account. There are no words suggesting the
interface-identifier needs to be stable.

>       interface-id         An opaque value of arbitrary length generated
>                            by the relay agent to identify one of the
>                            relay agent's interfaces

Any reaosn not to make this 32 bits? the IPv6 API already has 32-bit
interface IDs. Is there any reason to make it larger?

Thomas

_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg