Re: WGLC on the design draft

Francis Dupont writes:
>  IMHO the problem is simple: there are 3 definitions of load
>  balancing:

I do not agree that there is 3 definitions of load balancing. 

>  Charter's one:
>     o Load balancing. Multihoming is supported only in the sense of
>       failing over to another interface; sending traffic over multiple
>       addresses using the same SA is not supported.
>  (note this is the commonly accepted one)
> 
>  Section 3.2 one:
>        Note that MOBIKE does not aim to support load balancing between
>        multiple IP addresses.  That is, each peer uses only one of the
>        available IP addresses at a given point in time.
>  (note this is a very limited, and IMHO too limited)

I think those definitions are mostly same, the charter version does
not say anything what happens with multiple SAs, thus it keeps that
open. The charter definition still says that it is failing over to
another interface, which means it will not use multiple interfaces at
all, it uses one and if that fails, it fails over to another
interface, using only that after that.

>  Section 6.3 one which is not really described.

Section 6.3 does not define any load balancing, it simply says that in
future we might work on the load balancing issues too, thus we need to
make sure our protocol is such it can be extended. 

> What I'd like is a *place* for the limited load balancing (which is not
> commonly named load balancing outside the two current documents as it is
> *not* covered by the charter limitation) where a SA uses one address at
> one time but not all SAs use the same address.

You can already do that. It is in the protocol. Simply create multiple
IKE SAs and each of them can use separate address. Each IPsec SA
inside those IKE SA moves along with the IKE SA.

So we already have that feature you are asking for. It is also
described in the design document (last sentence of the section 5.3). 

> Now how to do this? 6.3 doesn't seem to be the right place because the text
> is clearly about possible address lists when what is needed are SPI lists.

6.3 is definately wrong place. The section 5.3 already talks about
list of SPIs and also list of ranges of SPI values. 

> So I suggest to fix the text in 3.2 and does not say that MOBIKE but the
> current design does not aim to support load balancing (keeping the same
> definition).

Load balancing in any forms is explicit non-goal for MOBIKE WG. The
charter says that very clearly.

The charter also lists that multihoming is only supported to support
failing over, not for other reasons. It also explictly says that
usingtraffic over mulple address for same SA is outside the scope, but
that does not mean it would be the only thing covered by load
balancing.

Load balancing in general is outside the scope of MOBIKE WG.

> Note I am not the only person who can compare the design document and
> the charter so IMHO this should be really addressed before the IETF last
> call.

I do not really find anything that would be conflicting between the
charter and the design document about the load balancing. Both of them
say load balancing is outside the scope of MOBIKE WG and the current
protocols. 
-- 
[email protected]