One architecture or interconnected architectures?
From FIpedia
The Disagreement
What is the disagreement?
Viewpoint 1
Summary
Viewpoint 2
Summary
Moderator
Stephen Hailes
Leave a Comment
RaeHarbird said ...
David Clark said ...
I think this disagreement needs to be "unpacked". It might be asked narrowly about architecture at the traditional packet level, where it might be seen as a question about the value of virtualization. But a future network will embody (or at least call for) common agreement at many levels. So we might ask about one packet format vs. several, or one information dissemination architecture vs. many, or one identity scheme vs. many.
In general, there has been a lot of benefit from having one packet format (it is a simple platform on which great diversity sits) but in general, we benefit when a design does not call for just one solution but permits many. Then ease of use can battle diversity of solution for the best outcome.
This preference, I think, is perhaps something to debate.
--Ddc 16:27, 21 June 2009 (EDT)
Dpapadimitriou said ...
Before launching this debate we should probably reach a bottomline consensus (common denominator) of what we mean by "architecture" in the networking context ? In brief, what's an architecture ? an architecture can imho be defined as the set of design principles from which (functional, informational, etc.) models are derived to characterize and specify its components (procedures, data structures, etc.) and their interactions (messages, calls, etc.).
Other opinions ?
--Dpapadimitriou 07:04, 11 July 2009 (EDT)
Dirktrossen said ...
So this is a disagreement in itself (what is an architecture)? I remember the meeting of the IRTF E2E WG in October 2005 where a similar question was asked. I still have a picture that shows the process of 'architecting a solution', which I developed during the meeting from the discussions taking place. We use an adapted version of the picture in the EU project PSIRP, which combines bottom-up and top-down approaches into an evolving lifecycle approach.
--Dirktrossen 11:52, 14 July 2009 (EDT)
Per-Olav Gramstad said ...
This looks like the question of system or the system of systems. The differentiation between the sum and its' parts only make sense once they are given interpretation.
In reality there are always several architectures, as there is always one being used, an old one being phased out and new one coming in.
I'm sorry but due to temporal characteristics, change, there can only be interconnected architectures ;)
--Per-Olav Gramstad 07:58, 2 October 2009 (EDT)

Sensors vs. servers
--RaeHarbird 12:04, 7 June 2009 (EDT)