[CII] End2End (was) terms and conditions

[tvest at eyeconomics.com]

On Nov 30, 2009, at 2:41 AM, John Osmon wrote:

> On Sun, Nov 29, 2009 at 11:54:51PM -0500, tvest at eyeconomics.com wrote:
>>
> [...]
>> The failure of any single, enterprise-specific payment mechanism
>> (e.g., a department store credit card), or even of a single bank can
>> be a major problem for the its current customers/users, employees,  
>> and
>> investors. But thanks to the flexibility and basic inter- 
>> institutional
>> neutrality of the monetary economy, customers can move their deposits
>> elsewhere and find other ways to send and receive payments, workers
>> can find other places or means to earn money, and investors can find
>> other opportunities where their capital resources can yield more
>> stable returns.  So long as that capacity for dynamic readjustment is
>> preserved, no single element or arrangement of elements within the
>> monetary economy is sacrosanct, and no single atomic or aggregate-
>> level element within the system would necessarily qualify as a
>> "critical infrastructure." Given that flexibility, the overall  
>> pattern
>> of system elements can (and perhaps should) be decomposed thoroughly
>> from time to time.
>
> I like the gist of the above.  I read it as:  "Critical economic
> infrastructure is intact if individual institutions can communicate."
>
> Bill Manning points out that "pairwise" BGP relationships allow for
> periodic tests of connectivity.  Is it sufficient to say that
> economic/banking crictial infrastructure is intact if the majority
> of those institutions can continue to "talk" during a crisis?

Hi John,

Thanks for the response.
Although I wouldn't say that it's erroneous, I don't think that I  
would embrace this particular summary reformulation myself.
Perhaps it works best as a description of the perspective from any  
single atomic or aggregate-level system element (i.e., an individual,  
an enterprise, a jurisdiction, et al.), but even then to me the  
characterization "intact" seems excessively static and binary. IMO,  
e2e is important because it recommends a distinctive form of inter- 
process signaling/interaction that critically affects both  
communications within the system as it exists at any point in time,  
and also the responsiveness of the system to the addition (or  
subtraction) of individual components over time. While a system might  
be perceived as "intact" by individual participants so long as it  
supports their own (bilateral) interactions, if the system's e2e  
characteristics are degraded in ways that would impact interactions  
that they (or others) might wish to have in the future -- including  
with functions that might not even have been incorporated into the  
system yet -- then in fact the system is no longer "intact," even if  
some (or even most/all) participants have not yet recognized that  
fact. In this sense, e2e is a description of a dynamic system with an  
open frontier, or an unbounded future. It seems like your  
reformulation would be consistent with that kind of system, but would  
also work in systems that are robust to failure although not  
necessarily open to more elements, including completely novel elements.

> If so, we need to increase the "splay" of networking between the
> various economic/banking entities that appear to be critical.  If we
> can keep enough pairwise connections up, "dynamic readjustment" is
> preserved.


In my own formulation, the idea of "splay" as a feature that  
individual system elements (including aggregate-level elements like  
institutions) may possess or not possess is obviated by the more  
fundamental flexibility/adaptability that is an inherent feature of an  
e2e-defined systems. By definition, an e2e-defined system permits such  
"dynamic readjustments" whenever and wherever they are required/ 
preferred -- but that does not necessarily guarantee that such  
adjustments will always happen. Thus, whenever some specific aggregate- 
level system participant is unable or unwilling to make a necessary  
adjustment, the aggregate may cease to exist, but its constituent  
elements (e.g., the customers, employers, and investors in my previous  
example) can remain integrated into the broader system (albeit in  
different roles or capacities), and the system itself endures.

Granted, this may not be especially comforting to enterprises with CI  
concerns, but I imagine it might at least suggest some potentially  
useful forward-looking strategies.
Perhaps something along the lies of "enhanced awareness of splay  
opportunities that are inherent in the system"... In fact, that sounds  
like the sort of thing that Barry might have had in mind when he  
raised the e2e question... (?).


> If the above is true, can we generalize this approach to other
> critical functions?


Actually, even if the above is not true, I would say that the  
generality of this framework is already quite explicit, at least in my  
own work.

Perhaps I can make that point clearer by clarifying the relationship  
between the above and the canonical e2e arguments. Consider: If fate  
sharing is the core idea behind e2e, and the system to which we would  
seek to apply e2e is understood to encompass multiple parallel,  
intersecting, and overlapping functional elements, then fate-sharing  
necessarily implies four distinct kinds of potential problems: a  
functional component's degree or level of effect that is either (a)  
excessive or (b) inadequate, and the scope of those effects with  
respect to (1) the function's intended target scope, and/or to (2)  
other, unrelated, out-of-scope system components. On this view, e2e is  
a particular strategy for system design in which functional features  
are modularized, scoped, and arrayed (in some cases as a result of  
evolution/trial-and-error as much as by conscious design) in such a  
way that, in the aggregate, the system delivers a mix of rank-ordered  
benefits (or solves a mix of rank-ordered problems), in which the  
ordering of goals/problems roughly corresponds to the placement of  
functions along some notional spatial and/or temporal axis ("notional"  
b/c of the system's more complicated dimensionality). In the specific  
case of the system of TCP/IP-based addressing and routing, the  
function that enjoys highest priority is that of (mere) attachment,  
including re-attachment -- a.k.a. resilience, adaptability -- as well  
as new, first-time attachment -- a.k.a. system openness, including to  
completely novel, "unprecedented' elements. By design, that goal  
trumps the legitimate but subordinate goal of (reliable, "quality")  
attachment, et al.

This combination of adaptability and openness, which distinguishes the  
Internet from both its constituent inputs and from its predecessor  
technologies, exactly parallels the features that distinguish a  
"monetary" economy (i.e., a heterogenous, decentralized exchange  
system in which diverse bilateral transactions are dramatically  
simplified by the use of a common mediating technology called "money")  
from the sort of ad hoc, barter-based economic systems that predated  
the discovery of the advantages of technologically-mediated exchange.

Put more succinctly, the system of monetary instruments and financial  
flows and the system of TCP/IP addressing and routing are at root just  
different implementations of the same critical function: they're both  
what monetary economists (esp. those that specialize in the subfield  
of monetary "search and matching") would call "liquidity mechanisms."  
In fact, if you take into consideration the two systems' basic moving  
parts, and all of their various bilateral relationships, flows,  
motivations, and then consider the primary existential risks to which  
both systems are vulnerable, and even the most common risk mitigation  
strategies that have been adopted in both systems, you find that the  
two systems are almost perfectly isomorphic.

Hard to believe at first blush, I know, but the supporting evidence is  
extensive and growing, with no confounding observations in evidence to  
date despite close to three years of skeptical scrutiny, including by  
quite a few of the most experienced protocol designers and operators  
around. I'm working on an introductory article on this now... can  
provide more information in the interim if you're interested.

Regards,

TV