[e2e] Just as an idea. Why can't we use hop by hop flow control for TCP?

Detlef Bosau detlef.bosau at web.de
Tue Mar 11 13:04:56 PDT 2014

Am 10.03.2014 09:50, schrieb Roland Bless:
> Hi,
> On 06.03.2014 17:58, Detlef Bosau wrote:
>> I would like to add a question mark here.
>> TCP is no way "only running at the end points" - of course, links and
>> switching nodes along the path are involved and affected.
> They are usually not holding TCP state.

And neither they do in my concept. They would hold flow control state -
in a manner which would reasonably scale up to a huge number of flows.
>> The more I think about it, this appears artificial to me.
> No, look DCCP, for example, provides CC but now flow control, because
> it is unreliable and you don't care whether the receiver must drop
> packets - for real-time applications this is advantageous because
> they can try to catch up.

To my understanding, DCCP takes TCP congestion control and reimplements
this for TCP flows.

So, it basically suffers from the same weaknesses.
>> I wrote this sometimes before: When you think so, your model is:
>> Sender -------------------------Line---------------------- Receiver.
> No, it's not the model I have in mind...but it nevertheless depends
> on the level of abstraction and layer(s) that you are looking into.

It is the model used in the congavoid paper.
> You can easily have models that take all such things into account,

Hm. The very challenge is to take the right things into account and
leave the others.
>> I further think that this is a misinterpretation of Salzer's paper.
>> Following Salzer, we should solve local problems locally and global
>> problems globally. Actually, we solve problems at the end points. (If at
>> all.)
> Are you referring to the end-to-end argument paper?

Of course.
> IMHO one of the essential conclusions is that you often need
> application knowledge in order to behave adequately in case
> of failures 

I agree - however, the applications may well decide themselves - given
the appropriate information.

> and that you better avoid to put this knowledge
> (or state) into the network infrastructure. In the e2e argument
> paper there are no local/global scope aspects mentioned.

So they are forbidden?

Roland, when was the e2e argument paper written?

Correct. 1984. When I went to supermarket this afternoon, the "Internet"
constituted by the iPods there was greater than the whole Internet at
the time of Saltzer's paper.

Again you will tell me that sounds to harsh.

Does it cause harm to anybody when I consider an alternate way?

> Roland

Detlef Bosau
Galileistraße 30   
70565 Stuttgart                            Tel.:   +49 711 5208031
                                           mobile: +49 172 6819937
                                           skype:     detlef.bosau
                                           ICQ:          566129673
detlef.bosau at web.de                     http://www.detlef-bosau.de

More information about the end2end-interest mailing list