[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ih] protocol structure -> economic structure

At 11:43 +0000 2012/03/06, Tony Finch wrote:
>Craig Partridge <craig at aland.bbn.com> wrote:
>>  Dave Clark has been giving speeches for about a decade on this issue.
>>  He has a 2002 paper entitled "Tussle in Cyberspace" and the term
>>  "tussle space" has some traction for discussions about the implications
>>  of particular protocol designs.
>I think that's probably the paper I was struggling to remember, thanks!
>I've read John Day's book as well, though more recently. I think
>After a quick look to remind myself about Clark's paper, it's very much a
>manifesto for the future internet architecture projects, and about
>designing architectures to fit economics and politics. The reason I failed
>to find the word "architecture" when writing my original message was that
>last week I had been thinking in the opposite direction. That is, given
>a (set of) protocol(s), what is the implied architecture? What economic
>structures are likely to arise?

Having been involved in a lot of these economic wars, some how I 
always found "tussle" to be a bit wimpy for describing them.  ;-) 
What I witnessed was a whole lot more intense and had a whole lot 
more at stake (at least the participants thought so) than a mere 
tussle.  ;-)

>For instance, the DNS's multi-component names are (in the protocol) always
>given as rooted, absolute paths, which implies a tree structure and
>therefore a hierarchial organization, and paternalistic / authoritarian
>politics. But if you make the names relative, that implies a rootless
>graph structure, a flatter organization, and a more libertarian /
>anarchist politics.

Interesting, not my interpretation at all.  Since duplicate 
assignment of names must be avoided, I always saw a hierarchical name 
space as the minimal organization to meet the requirement:  Give an 
organization a branch and let them do whatever they want.  And a 
deeper tree allows more local uses of a name to have a common usage, 
e.g. more than one JoesPlumbing in different part of the tree.

Remember with computers, relative names just means you have 
established a context in which you can ignore the whole absolute 
name.  As Utah Phillips defined it, "An anarchist is someone who 
doesn't need a cop to tell him what to do!" ;-)  Or my personal 
favorite, when asked by an interviewer, Karl Hess, a founder of the 
Libertarian Party who lived on barter for 30+ years (long story 
there) and a friend said "An anarchist is a good friend, a good 
neighbor and a good lover."  The interviewer asked, is that all? 
Karl said, "What do you want?! Rules!?"

Self-organization requires a minimum of rules.  Hierarchy allows that.

Now that said,  ICANN seems to have found a way to make it a whole 
lot more shall we say "interesting."  ;-)

>But note! The politics has more effect on who the architecture appeals to
>than the influence it has on society. I think we have a habit of
>attributing the culture of the early adopters of a communications medium
>to the medium itself, when in fact communications tools are much more

I would agree that the communication tools should be neutral.  In 
what we have been pursuing, we have been looking very carefully at 
the structure of the problem, rather than the economic or political 
environment.  Interestingly enough, concentrating on the nature of 
the problem leads to some interesting implications.  For example, a 
realization that a global name space is in fact relative.  It has 
been interesting to see how our habits of thought have lead us to 
assume things that were not the case.

The interesting thing I have found is that when attempts are made to 
turn "architecture" to economic and political ends that cross the 
invariances inherent in the problem, things get very messy.  We see 
this with the "beads-on-a-string" model, we saw this in the OSI 
model. (The problems in the Internet model (as much as it can be 
discerned) have more to do with not recognizing fundamental concepts 
than trying to impose political solutions.)

The original topic I thought was economic impact. That is very clear. 
Early on in the ARPANET we often talked about the political impact. 
(It was the time!) ;-)  However, if anything the impact has not been 
quite what we expected.  If anything it has lead to balkinzation more 
than anything.  Although, it is not alone in doing that.

>Perhaps I'm taking this analogy too far. Another example (much more
>grounded in reality) was Dave Crocker's description of the Internet Mail
>architecture. Email has not been without architecture (the 821/822 split;
>X.400 but it has also been shaped a lot by external forces such as interop
>difficulties (late 1980s / early 1990s) and spam (which drove people to
>implement the architectural distinction between message submission and
>relay, amongst other things). So I suppose it would be informative to ask
>to what extent tussles moulded email, and to what extent email shaped
>those tussles.

First one would expect application protocols to tread more into this 
realm than the pure networking, or what I would call IPC.  The 
application protocols that interact with humans should reflect this 
more than others.

As for the email example, the distinction between submission and 
relaying was not in response to spam but much much earlier, once we 
had hosts (PCs and workstations) that were not connected all the 
time.  (There were also networks in this category, but I will 
distinguish them.)

I do not consider email "relaying" to be relaying in the same sense 
as relaying IP packets.  Relaying as an IPC mechanism is constrained 
(as all IPC is) by the requirement that MPL be bounded. (Maximum 
Packet Lifetime) This is an impiication of Watson's results.  This is 
not the case with email.  Email is remote storage.

This is one of those interesting cases, where how it appears to the 
observer and how it appears to the organism (in this case the 
network) are quite different.  Another case where science disproves 
our intuitions   Which of course why we do science.

Take care,