Re: [P2Prg] CORE subgroup problem statement

[Arnaud Legout <Arnaud.Legout at sophia.inria.fr>]

Hi Luca,

Luca Caviglione wrote:
> I do agree here. Other is a better name and also a weak classification. However,
> I think that correct classification (better, the one I would prefer) must rely
> on architectures, rather than functionalities.
>   
I am not sure I understand why you want to focus on architecture. As you 
said below, that only gives you information
on how peers are organized. They what can you do with that?
A taxonomy should have an intent, when it is based on functionality it 
offers a classification of building blocks that can be
used to design new protocols.
At least, I would have understood a refined taxonomy based on 
architecture. But the taxonomy structured/unstructured
is too coarse to be useful.
> Actually, BitTorrent is a superposition of two protocols. The BT-Tracker
> protocol and the BT protocol for the swarm. Then, the protocols reflect the
> architecture where they act. However, this is my perspective and it doesn't
> mean it is the best one or better than others.
>   
Now, I see why you classify it as hybrid, but here again I do not see 
how this classification is useful.
I expect from a taxonomy to be able to compare protocols within the same 
classification.
But, protocols classified as structured, unstructured and hybrid are so 
different, that I cannot really compare them.

> > <AL>
> > I am not sure it is possible to abstract the architecture from the
> > technique..Unstructured is not specific to P2P as well. Multicast has
> > been using this kind of architectures and techniques (flooding,
> > expanding ring search, etc.) for a long time.
> >     
>
> I partially agree. However, I think that the first step, in order to produce a
> taxonomy, is to separate causes with effects. You can use swarming, hording (or
> flooding and expanding ring) since the architectural blueprints allow such
> techniques.
>   
I am not sure what you  mean by cause and effect. If you mean 
architecture and functionality, I am not sure
that starting with architecture is the right first step.
My argument is that in the context of P2P the architecture, i.e., how 
peers are arranged, does not give you much information.
I agree it is better than nothing, but a taxonomy based on functionality 
would be more useful.

> For me, structured means that there is an enforcement in the organization of
> peers. Having the peers forced to arrange in a ring (as an example) allows to
> exploit some search strategies. On the contrary, having peers connected
> randomly, brings you in the case where you don't know where the content is. So
> you must flood or something else.
>   
I see, but you can have peer structured in a ring for content 
localization or for file replication (I agree this is not the best
architecture for file replication, this is just for the example).
The case of unstructured is more realistic. You can have a random graph 
connectivity for content localization or for
content replication.
Thus, knowing that your protocol is classified as structured or 
unstructured is not of much help.

> ...and probably is there the main knot. For me, unstructured and structured is
> how "the peers are organized" and not related to localization techniques. I
> mean, if I am going crazy, I can use a flood based technique selecting randomly
> peers in Chord's finger tables (and probably I can patent this as the useless
> search system ever).
>   
and that is my point (I am starting to be very redundant), the taxonomy 
based on structured and unstructured does not give you
a very useful information.

Regards,
Arnaud.

_______________________________________________
P2prg mailing list
P2prg at irtf.org
https://www1.ietf.org/mailman/listinfo/p2prg