Re: [Masque] FW: New Version Notification for draft-ihlar-masque-datagram-numbers-00.txt

[Mirja Kuehlewind <[e-mail]>]

Hi Christian,

the IAB is working on guidelines:

https://datatracker.ietf.org/doc/draft-iab-path-signals-collaboration/

Adding the option to provide re-ordering to MASQUE is inline with these guidance as we give the client the choice to request this service or not. Following these guidance you don't need to "undeploy" anything because you simply just don't request it. Also "undeploying" existing middleboxes might be a matter of time due to more and more encryption. I don't think we need a separate strategy for that, beyond what we are already doing, namely driving the deployment of encryption and at the same time work on new collaborative approaches for network support (as we do with MASQUE).

Mirja



On 29.10.22, 21:31, "Masque on behalf of Christian Huitema" <[e-mail] on behalf of [e-mail]> wrote:

    On 10/29/2022 12:27 AM, Mirja Kuehlewind wrote:

    > not all traffic on the Internet is QUIC or TCP, so you simply don’t know if the end-to-end handles re-ordering well. However, this is not only about in-order delivery. TCP flows without RACK will misinterpret high re-ordering as congestion slow down its sending rate to minimum. Fortunately, re-ordering is handled well by modern transports, so in-network re-ordering should not be the default anymore. There are still cases where it’s useful.


    Mirja,

    You are providing one of the classic arguments for "putting smarts in 
    the network", i.e., that there are some kind of inadequacies in some 
    variants of endpoints. This is of course often used to justify deploying 
    middleboxes. But then, we know how that evolves. The middle-boxes may in 
    some case provide performance gains in the short term, but they always 
    end up lowering performance and creating ossification in the long term. 
    Maybe that's a problem the IAB should study: how to manage middlebox 
    deployment so that they only affect those devices that benefit from 
    their services; how to "undeploy" these boxes when they cause more harm 
    than good. Or maybe issue a recommendation that such things should not 
    be deployed at all in the first place.

    -- Christian Huitema

    -- 
    Masque mailing list
    [e-mail]
    https://www.ietf.org/mailman/listinfo/masque