Re: [netext] second revision of the new charter for theworking group

[Carlos Jesús Bernardos Cano <[e-mail]>]

Hi Jari,

On Mon, 2010-01-25 at 23:49 +0200, Jari Arkko wrote: 
> Pierrick,
> 
> > Actually, the "single interface" approach may not be the only way. For instance, draft-bernardos-mif-pmip leverages on the weak host model (RFC 1122) to address IP flow mobility issue and leave the host unmodified. This is why we think the charter should not be so "solution oriented".
> >   
> 
> Lets talk about this for a second.
> 
> It seems that there are a number of different host impacts.
> 
> 1) We might need a new protocol between the host and the network. This 
> implies new code on hosts and network and new protocol messages on the wire.
> 
> 2) We might need just new code on the host, but still be able to 
> determine all necessary information from existing exchanges.
> 
> 3) We might need to assume that the host implements something which is 
> allowed according to the RFCs but has not been implemented.
> 
> 4) We might need to assume some code on the host that is often but not 
> universally there on all current hosts.
> 
> 5) We might need to assume some configuration on the host.
> 
> Obviously, on a network where someone is in full control we can assume 
> anything, and build any amount of new protocols and code to optimize the 
> behavior. Such networks are VERY rare, however. For instance, I'm 
> writing this mail on my laptop which is connected over my cell phone to 
> 3G network, and the network can make no assumptions about what features 
> my kernel has on my laptop. Or what my config is. Yet, a phone or a USB 
> stick that implements 2G to 3G switching works just fine with this 
> configuration.

Well, I partially agree with this. Your laptop works OK with that
configuration, but if in addition to the USB stick you add a wifi
connection, then your laptop may not work just fine without assuming
something else. Besides, if you want to benefit from sending http
traffic through one interface (e.g. wifi) and voip traffic through
another one (e.g. cellular), then you need something else (some
support/configuration at the laptop). If we want to enable flow mobility
in nodes with multiple interfaces in a PMIP domain, we may need to
assume that the mobile node is at least capable of doing some things
(where capable may imply certain configuration/standard support). A
single IP interface sending/receiving over multiple media (flow mobility
not just vertical handovers) will need to know "how" to dispatch flows,
something that is not implemented in host devices yet (at least widely,
as far as I know).

Carlos

> 
> My point is that we are not after no host impacts in some theoretical 
> standardization sense. We actually want devices to work, old devices, 
> current devices, new devices. Out of the box. And whether they work is 
> affected by a number of things, including lack of code or config. I 
> don't feel any happier about my hosed connection if the RFCs would have 
> allowed it work. Or if the code is there but I didn't know I needed to 
> turn it on. If it doesn't work, it doesn't work.
> 
> Jari
> 
> _______________________________________________
> netext mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/netext

-- 
Carlos Jesús Bernardos Cano     http://www.netcoms.net
GPG FP: D29B 0A6A 639A A561 93CA  4D55 35DC BA4D D170 4F67