Re: [RRG] Are host-stack modifications allowed or disallowed ?

[Luigi Iannone <[e-mail]>]

Le 06-mars-08 à 00:03, Dow Street a écrit :

> On Mar 5, 2008, at 11:10 AM, Tony Li wrote:
>
>> |> Yes, but there's a very big difference between you driving your  
>> car
>> |> the way
>> |> that you want and the way that hop-by-hop forwarding works.
>> |
>> |What if the driver could (optionally) select from a set of back  
>> roads
>> |that is limited to those published by the network provider?  If the
>> |driver doesn't want to and/or cannot do this additional navigation
>> |work, he could follow the nudge of the underlying hop-by-hop  
>> mechanism.
>> |
>> |In other words, maybe we can build a basic reachability service of
>> |global scope, and layer on top of that a more advanced service that
>> |provides enhanced functionality, but where the costs are localized.
>>
>>
>> That would be a fine thing, but the user has only two ways to  
>> express his
>> desires: one is by injecting routing information, which  
>> necessarily can be
>> overridden.  The only other way is by doing a setup protocol.
>
> In the model I'm thinking about, the network (i.e. provider) would  
> publish a set of path options (e.g. a map, list of potential  
> waypoints, set of deflection bits, etc) for the host to consider.   
> The host could then choose from among these options, with the  
> expectation that its choice will be honored by the network in the  
> general case.  Otherwise, the network should change the set of  
> options it is publishing.  The provider retains control by setting  
> the bounds of the allowable routing policy.

You should ge a look to:
https://datatracker.ietf.org/drafts/draft-bonaventure-informed-path- 
selection/
https://datatracker.ietf.org/drafts/draft-saucez-idips/

they both talk about a way for ISP to suggest path to ned-host.

Luigi


>
> The publication of path information from network to host might be  
> considered a setup protocol, but I'm thinking it would be strictly  
> optional.  The host could still send "regular" packets without  
> indicating a specific path preference, and get whatever service the  
> network gives them.
>
>
>> Some folks have proposed a hybrid architecture in the past that  
>> allowed
>> both, but it (rightfully) went over like a lead balloon.
>
> Does the model described fit your definition of a hybrid architecture?
>
> R,
> Dow
>
> --
> to unsubscribe send a message to [e-mail] with the
> word 'unsubscribe' in a single line as the message text body.
> archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg

Luigi Iannone


[e-mail]