ch7-conclusion.tex

\chapter{Managing future routing table growth}
\label{chap:conclusion}

This thesis set out to investigate whether the CIDR Report was effective in
motivating collective action to mitigate the effects of routing table growth,
and to draw conclusions about how to manage the Internet routing table. By
analyzing the route announcement and aggregation behavior of ASes after
appearing on the CIDR Report and comparing this to the behavior of ASes that
never appeared on the CIDR Report, we found that an improvement in AS
aggregation behavior was likely associated with appearing on the CIDR Report,
and that this treatment effect decreased in more recent time periods. The
overall routing table continued to grow in spite of this treatment effect,
though this was likely due to the organic growth and participation of new
networks, rather than continued deaggregation of CIDR Report participants.
Regardless, the routing table has not recently grown at a rate that exceeds
upper bounds of our technology (Figure \ref{fig:li_router_scalability}) or the
capabilities of routers available on the market. We now turn our attention to
second objective of the thesis, about how to manage the Internet routing table.

% \section{Returning to the problem of routing table growth and deaggregation}
% With this summary of our observations and lessons learned from studying the
% CIDR Report, we return to the broader issue of this thesis regarding
% collective action and the sustainable growth of the routing table.

\section{Is routing table growth a problem?}
The first question that must be asked in considering the issue of managing the
routing table is whether routing table growth is even still a problem affecting
the scalability of our current interdomain routing architecture. The Internet
routing table is no longer the driving force behind router capabilities
\cite{Davie:2011uq}, and as illustrated in Chapter \ref{chap:intro} and also
discussed in \cite{Fall:2009fk} and cited by \cite{Huston:2011ys}, Moore's law
has outpaced routing table growth.

However, there are also evolutionary changes taking place in the Interdomain
routing space that may cause the Internet routing table to grow faster than has
come to be expected. The Internet will likely be using IPv4 addresses for some
time, given the slow pace at which the transition to IPv6 has taken thus far.
While the forthcoming exhaustion of free address space will likely hasten IPv6
adoption, it will also likely incentivize the more efficient use of IPv4
addresses in order to extend the useful lifetime of IPv4 before IPv6 adoption
is deemed complete. This more efficient use will likely be accomplished through
partitioning larger address blocks for trading, as well as the deaggregation of
larger blocks into smaller, more numerous announced prefixes to facilitate
expression of routing policy for larger blocks of machines behind NATs and
other transition technologies. Both of these strategies will result in
increased numbers of IPv4 prefixes in the routing table.

At the same time, as the IPv6 routing system moves from experimental and
prototype implementations to production-ready networks that more closely mirror
the IPv4 routing system \cite{Cowie:2010vn}, the number of prefixes in the IPv6
routing table will also grow to be larger than present, and indeed currently
appears to be growing exponentially (though the table itself is still quite
small\footnote{The IPv6 routing table contains approximately 6000 prefixes, but
currently appears to be growing exponentially:
\url{http://bgp.potaroo.net/v6/as6447/}}). If IPv6 is deployed via dual-stacked
routers, as is often the case, this means that available router resources for
storing routing tables and processing BGP updates will be contended for by two
routing tables instead of just one. Eventually the IPv4 table should be able to
be deprecated, but it is unknown when that will be viewed to be commercially
viable---a point at which all customers have at least as good IPv6 connectivity
as IPv4 connectivity.

So while routers currently have sufficient capacity to meet ISP internal needs
while also adequately supporting the Internet routing table, the scaling
constraints of BGP that were introduced at the beginning of this thesis have
not been eliminated; Moore's law has simply allowed these limitations to be
outpaced by technology. If the growth properties of the Internet routing table
do change in future, our interdomain routing system may again be strained or
made less robust by its own growth, as there is nothing that ultimately
constrains the size of the routing table.

If we agree that there are at least potential problems of scalability or
robustness affecting our interdomain routing system under certain routing table
growth scenarios, the next question is: what should be done about this problem?
Assuming we believe the Internet is too valuable to allow to become unreliable,
then some action must be taken. A solution space is outlined below, focusing
on uncoordinated solutions (solutions that can be implemented unilaterally) and
coordinated solutions (solutions that require collective action). Both technical
and non-technical solutions are discussed.

\section{Uncoordinated solutions}
The first set of solutions that we can consider to this problem are
uncoordinated solutions---measures that require only individual action to
enjoy individual benefit in terms of mitigating the effects of routing table
growth. Solutions in this class are easy to deploy because they do not require
the collective action that a protocol transition would entail. However, they
are also somewhat limited in that they must work within the existing technical,
business, and policy architecture of the Internet and its interdomain routing
system. In other words, these solutions cannot change the underlying scaling
properties or route aggregation incentives of our current system, and instead
use other tactics.

In this class of solutions there are several technical approaches that have
been proposed, as discussed in the related work. Generally these involve
spreading the routing table across multiple routers (virtual aggregation)
\cite{Ballani:2009tg} at the cost of adding additional hops and packet
processing latency, or performing aggregation of the RIB inside the router. The
latter approach, which is similar to the aggregation performed by the CIDR
Report (though almost certainly using a different criteria for the ``same
routing policy'' invariant, such as next-hop IP address),
% also, ACLs will affect aggregability -- whooops
is a solution preferred by network operators \cite{Zhao:2010fu}, and could
certainly improve the longevity of the more expensive FIB routing table slots
for individual networks that deploy this solution. Further, it is likely that
the operators affected by the size of the routing table, such as DFZ network
operators, would implement a solution like this (providing their vendors
supported it), allowing other networks to continue to rely on them as well. The
in-router aggregation approach, if applied with similar results to the CIDR
Report could reduce the size of the FIB by approximately half\footnote{The
ratio of prefixes in the fully aggregated table to prefixes in the current
routing table is 175224/355045 $\approx$ 0.49 as of 28 January 2011, using our
implementation and data sources.}. Table growth induced by IPv4 transfers and
IPv6 deployment would not typically be aggregable, but this approach could make
more room for these new sources of prefixes by reducing the size of the routing
table due to deaggregated prefixes.

Also in this class of uncoordinated solutions are non-technical solutions.
Unlike the technical solutions which mitigate the effects of routing table
growth, these are proposed to work by creating incentives for announcing fewer
or more aggregated routes. One option in this space is to continue using the
CIDR Report as it exists today. The CIDR Report is included here, instead of the
following coordinated solutions section, because of the acknowledgment that it
is not a CPR governance institution on its own, and so without a strong
institution using the information it provides, it can at best inspire
individual action to persuade others or induce oneself to reduce their impact
on the routing table. Given that the CIDR Report appears to be ineffective now,
both from operators' perspectives and from our analysis, it could possibly be
improved based on the observations Chapter \ref{chap:discussion} in hopes that it
might better activate the latent norms-based behaviors that seemed to make the
CIDR Report effective in the past. However, this does not appear to be an
extremely promising option given its recent history, operator attitudes towards
the report, the problems identified in Chapter \ref{chap:discussion}, and the
fact that the technical solution above would likely be more effective.

Bilateral economic solutions may also be possible. For example, a Tier-1
provider could begin to charge its peers and customers for the number of routes
they advertise. However, this would require all of the peers and customers to
acquiesce to this change instead of taking other action, such as switching
to a provider that does not charge for route announcements. The autonomous
nature and competitive forces of the Internet might thus make such bilateral
solutions untenable unless implemented by all Tier-1 providers simultaneously,
and so economic solutions of this nature are probably better considered in the
following section instead.

\section{Coordinated solutions}
Beyond individual technical solutions described previously, it appears that the
most promising solutions to managing the size and growth of the routing table
will require coordination and collective action to achieve. Whether technical
solutions that involve adoption of a new interdomain routing protocol, economic
mechanisms to settle the costs of propagating a route across the Internet, or
a CPR governance institution for interdomain routing, all of these approaches
will require the cooperation of most of the participants in the Internet's
routing system in order to be effective and realize benefit from the solution.

Regardless of the solution or approach, there are general challenges in
achieving collective action. First, as Olson \cite{Olson:1982ly} and others
point out, achieving collective action can be difficult when the benefits from
individual action are not concentrated, or are not selective, creating an
incentive for opportunistic free riding. Also, while not uncommon in other
spheres of life, the notion that the Internet is no longer composed of actors
with homogeneous interests is relatively novel, as identified by
\cite{Clark:2005rt}. Instead of a cadre of similarly-minded technologists,
there are now a much more diverse and potentially conflicted set of interests
at play in shaping the architectural evolution of the Internet, as well as
achieving their objectives through the architecture. Even CPR governance
institutions are not guaranteed to form in situations where they would provide
benefit to appropriators, as Ostrom \cite{Ostrom:1990fv} identifies, because
there is a constitutional collective action problem (a meta-problem) in
agreeing to develop an institution in the first place, before the institution
can then be used to resolve operational collective action problems.
Acknowledging that collective action will be required, and may be difficult, we
now sketch out some potential approaches in this space and ways in which the
collective action towards these ends might be achieved.

\paragraph{A new interdomain routing architecture}

The most common approach taken thus far when approaching the limits of an
Internet protocol or architecture is to replace the system with a new, more
scalable version (e.g.  NCP to IP(v4), IPv4 to IPv6, EGP to BGP, etc.). The
adoption of new protocols is challenging; this challenge increases when a
protocol is not backwards compatible with its predecessor, and increases even
more at the network layer of the protocol stack where the new protocol must be
coordinated along the entire end-to-end path between hosts, rather than just on
the hosts themselves. This dilemma often means that there is no benefit from or
incentive for incremental protocol adoption, thus making adoption difficult to
justify for rational, commercially-motivated actors. This, at least at a high
level, is plainly visible in the slow-moving adoption of IPv6
\cite{Cowie:2010vn} that has theoretically been ongoing for the past 16 years
\cite{rfc1883} and has only recently started to gain traction with the
coming depletion of unallocated IPv4 address blocks.

In this context, we can consider the adoption of a new interdomain routing
architecture and protocol such as one of the potential candidates described in
Chapter \ref{chap:relwork}. The move to a new routing architecture and protocol
involves many actors that are responsible for various components of the
Internet and the devices and software that connect to it: router and network
equipment vendors, operating system and application software developers, and
Internet service providers, as well as the organizations that purchase and
operate this hardware and software and purchase Internet service from ISPs. All
of these parties must coordinate in order to realize the transition to
a new architecture. Without clear incentives, such as has been claimed to be
the case with IPv6, the transition moves slowly until there is motivation
(address exhaustion in the IPv6 case or table growth scaling issues in the
routing table case).

In the case of a new interdomain routing architecture, the situation may not be
as discouraging as with IPv6. Not all routing architectures designed to replace
BGP are equally difficult to transition, and there may be other motivations and
incentives that would promote the adoption of of these replacement protocols
even without immediate and real routing table scaling issues. With regard to
transition difficulty, protocol designers have arguably learned lessons from
the difficult IPv4-IPv6 transition and have put more effort into considering
backwards compatibility and transition strategies. One interesting candidate in
this light is the Identifier-Locator Network Protocol (ILNP)
\cite{Atkinson:2010zr}.  It provides locator-identity separation utilizing
existing Internet technologies---DNS, BGP, and IPv6---in a backwards-compatible
way, while allowing for aggressive hierarchical aggregation following protocol
adoption. Further, there may be ``killer apps'' and new markets that provide
commercial incentives for deployment of new routing architectures that were not
similarly present with IPv6 \cite{Li:2011vn}, which has only recently been
motivated by the coming end of readily available IPv4 address space. For
example, the growing popularity of mobile devices and applications may motivate
the adoption of a routing architecture that splits locator and identity to
allow mobility and provide each device with a static identifier without the
added overhead of Mobile IP \cite{rfc3344}.


\paragraph{Economic solutions}

Economic solutions to the problem of managing the size and growth of the
routing table generally deal with imposing a cost on users of routing table
slots to provide an incentive for networks to aggregate and otherwise reduce
and manage their route announcements carefully to avoid incurring unnecessary
costs. These costs could be based on the scarcity of routing table slots and
market-based pricing to determine the value of each slot, or it could be
imposed arbitrarily by individual network operators. Regardless, the intent
would be to internalize the externality that ASes impose on other networks
through route announcement and especially deaggregation.

Economic approaches to managing the size of the routing table are orthogonal to
other approaches such as protocol changes. However, such solutions would likely
be most favorable in cases where routing slots are scarce, such as might occur
if a routing architecture transition failed and BGP remained. Economic
solutions have the advantage that they treat all participants equally, rather
than the unequal quasi-sanction for only the top 30 of the CIDR Report.
However, there are potential equity issues in an economic approach compared to
the current Internet. Depending on pricing of routing table slots, less wealthy
participants could be disadvantaged in participating in the interdomain routing
system. Further, a market-based pricing mechanism could allow the price of
routing table slots to be manipulated by concentrated interests for strategic
or anticompetitive purposes.

Proposals have been made for establishing bilateral contracts between providers
to carry routes upstream in exchange for compensation \cite{Rekhter:1997mi}.
The approach of bilateral dealing is probably best with the interest of
maintaining a decentralized Internet of autonomous networks.  However, for such
contracting mechanisms to be generally adopted, ISPs would need to collectively
agree or accept a change to the current settlement model used by ISPs. Such a
move would likely need to be coordinated amongst the DFZ service providers,
which some fear may raise concerns about antitrust and collusion
\cite{Li:2011vn}. Further, it is not clear that network operators currently
have an appetite for the complexity introduced by route carriage contracts
\cite{Zhao:2010fu}, though this could potentially change if faced with the
alternative of an overloaded interdomain routing system.

% - collective action required: need to change settlement approach in the market
%   collectively (antitrust?) or need to create a market -- cap and trade
% - added complexity of bilateral contracts in Internet operations
% - not clear that this is interesting to operators

% - maybe not "pricing" but just a mechansism where route advertisement is
%   compensated through contracts like those discussed in the literature
%
% - Assuming a loosely hierarchical structure (where you pay others to reach
%   prefixes and networks you can't reach on your own) customers would pay
%   providers to carry their prefixes to the greater Internet. Peers would pay
%   on the differences of prefixes exchanged between them.
%
% - these just provide a disincentive (is the gain from doing so bad?) -- they
%   don't actually prevent the routign table from growing excessively
%
% - if we absolutely need to control it -- cap and trade -- equity issues,
%
% - these treat everyone equally -- no disparity as shown in the CIDR Report

\paragraph{Governance institutions}

Much of this thesis has been spent discussing CPR governance institutions, and
indeed governance institutions are not often considered explicitly on the
Internet, though norms and community cooperation often play roles in operating
the Internet. More broadly, there is the option of developing an institution to
manage the Internet routing table. A central regulator of the interdomain
routing system would almost surely not be selected voluntarily by network
operators, and so ignoring the option of governmental intervention, we are left
with the CPR governance institution option. Again, like economic approaches,
the governance institution approach is orthogonal to the routing architecture
in place, but would likely be favorable only if interdomain routing remained a
problem because a protocol or architecture transition was infeasible or had
failed.

To develop a CPR governance institution, network operators would need to
collectively devise and commit to a set of rules governing use of the routing
table, as well as sanctions and monitoring mechanism (potentially like the
CIDR Report) to provide assurances of commitment. Having discussed many of
these issues in the context of the CIDR Report in Chapter
\ref{chap:discussion}, a routing table governance institution would need to be
more structured than the CIDR Report, and utilize collective choice and
sanctions, while also considering the need for multiple scales of operator
communities to enable better communication and social enforcement.

Further, there is the constitutional collective action problem (the
meta-problem) that Ostrom alludes to about how organizations come to agree upon
the rules and other characteristics of the governance institution that they
will then commit to. Ultimately, rational participants in the Interdomain
routing system would need to believe that they would be better off under some
institutional arrangement than the status quo. This would again mean that such
a solution would only be likely in the case where the robustness of the
interdomain routing system was at stake. However, in spite of the cooperative
norms of the Internet community, there is also sometimes libertarian sentiment
amongst operators in support of the belief that interconnection is voluntary
between autonomous networks. Establishing some form of collective order amongst
Internet operators may thus be more challenging than establishing norms of the
same nature.

% Interconnection is voluntary -- libertarian vs. collectivist thought

%- improve social/CPR governance?
%  - need to be able to answer questions about what is "fair"? ("know it when
%  you see it?") - would something like a routing table governance council
%  actually allow continued enjoyment of the dynamic aspects of internet
%  routing?
%  - need sanctions (see what else ostrom has to say about this)
%    - how do you actually do this? routing, but there is a problem of adoption
%    and market power and who has influence

% Sanctions are a challenge, and while there are some opportunities available
% within the routing system, the economic or operational ``damage'' that can be
% done is proportional to the content or customers you serve. Looking at other
% mechanisms for unilateral or multilateral sanctions within the Internet is a
% major opportunity for future work.

% This suggests that automated sanctions or pricing might be more effective.
%
% The challenge of agreeing to supply an institution that does anything more
% than provide information (i.e. sanctions) multilaterally is potentially
% problematic in looking beyond the CIDR Report.
%
% Interconnection is voluntary -- libertarian vs. collectivist thought

% * WHO ARE THE STAKEHOLDERS
% * WHAT ARE THE INTERESTS AT STAKE IN ACHIEVING THIS?
\section{Conclusions}

While the CIDR Report was once effective in encouraging aggregation of routing
announcements to manage the growth of the Internet routing table, it now
appears to mainly be a historic artifact of the previous scalability challenge
facing the interdomain routing system that led to the the adoption of CIDR. As
such it served a different purpose and focused on different behavior than may
be problematic in the interdomain routing system today. However, it is
noteworthy that it appeared to influence operator route aggregation behavior
for some time even after the CIDR transition, as both claimed by operators and
shown by the analysis in this thesis.

If interdomain routing does face growth-based scaling concerns again, is it
likely that uncoordinated action in the form of in-router aggregation will be
adopted to provide a short-term solution, while a longer-term solution is
developed in the form of a new, evolved routing architecture. Indeed, this is
the approach recently recommended by the IETF \cite{rfc6115}, with the
recommendation of the incrementally adoptable ILNP as the protocol for future
study and development as BGP's successor. While ILNP is likely more easily
deployed than IPv6, the coordination necessary to transition to a new
interdomain routing architecture will still be difficult and incentives should
be considered in hopes of avoiding the delayed transition currently being
experienced by IPv6.

Given that a routing architecture transition that changes the scaling
properties of interdomain routing will ideally eliminate the need for economic
or institutional approaches to managing the Internet routing table, a technical
solution is arguably preferable in this specific case.  However, governance
institutions in the style of Ostrom's CPR institutions should not be ignored or
disregarded generally as a solution to other collective action problems facing
the Internet, especially those that do not have obvious technical solutions
like the interdomain routing system does.

For all of the critiques leveled against the CIDR Report as a governance
institution, its partial efficacy is a testament to the social forces---the
strength of norms and importance of reputation---in the Internet operations
community. As network operators and others who rely on them work to solve the
difficult problems facing the Internet that are not purely
technical\footnote{Problems that may be susceptible to social mediation include
spam email, interactions with malicious or untrustworthy actors, or intentional
``cyberattacks'', etc.}, the leverage of inter-ISP relationships and the forces
within the social network of the Internet operations community should be
considered. Whether through the development of CPR-like institutions, or as a
part of larger and more comprehensive solutions, the social forces within the
Internet operations community may be effective in cases where purely technical
solutions have previously failed.

% - who would lead this?
%     - people who suffer -- Tier-1s/DFZ/etc?
%     - perhaps these people could come together for a partial
%       solution?

% \begin{itemize}
%
% \item{technical solutions in the short term -- driven by short-term pain or
% forward thinking}
%
% \item{protocol replacement in the long term -- likely driven by other benefits
% or pain (as with CIDR deployment), though more difficult to coordinate}
%
% \item{coordination will be difficult regardless -- CIDR is relatively
% miraculous by comparison and so the IETF/IAB, router vendors, and operator
% communities will need to think about how to execute our next transition
% effectively and with the proper incentives, hopefully learning lessons from
% the difficulties of IPv6.}
%
% \item{governance institutions would be useful in solving a number of the
% problems of the Internet, and perhaps more efficiently, but may be difficult
% to scale and coordinate Internet wide, and possibly also ideologically
% difficult}
%
% \item{CPR institutions still hold promise -- a call to action for these people
% to consider that or allow it to function within}
%
% \item{RIRs serve a good ``nucleation point'' as a forum where providers can
% organize into regionally-scaled CPRs}
%
% \end{itemize}

\section{Directions for future research}
This thesis on the CIDR Report and social forces at play in the Internet
operations community have exposed a rich area for further study in
understanding the effects of the CIDR Report and the nebulous underlying
governance mechanism that enabled these effects. In terms of routing table
growth more broadly and Internet collective action and governance institutions
more broadly still, the use of governance institutions as an alternative to
other proposed solutions for collective problems facing the Internet could use
more study and exploration about potential opportunities therein.

With regard to this work, there are several methodological improvements that
should appear in future iterations of this work that would greatly improve the
quality of the analytic process and hopefully reduce a number of the
confounding factors identified previously. First, measurement of the
pre-treatment behavior of ASes that would eventually appear on the CIDR Report
would allay concerns that some behavior other than appearance on the CIDR
Report was causing changes in aggregation behavior. Perhaps more importantly,
construction of a more representative control group (such as the ASes who are
just shy of appearing on the CIDR Report) would afford greater confidence that
the observed behavior change following a CIDR Report appearance was not due to
some of the biases mentioned in Chapter \ref{chap:analysis}. Finally, it would
better to utilize a statistical test, such as a Kolmogorov-Smirnov test, to
determine whether the treatment and control behaviors actually differ
significantly.
% Finally, performing a sensitivity analysis on some of the assumptions and
% design decisions built into our analysis would expose the

Beyond these methodological improvements, there are also a number of further
analyses and new directions to explore regarding the CIDR Report. Improving our
analytical tools to allow more fine-grained analysis of CIDR Report response
over time and amongst classes of ASes (i.e. ISPs vs. others) could expose other
behaviors not observed in this analysis. Our observation about the sensitivity
of the CIDR Report to variations in prefixes received from peers at the vantage
point may also merit further study to estimate the degree to which outlier
prefixes have contributed to inflated rankings on the CIDR Report.  Finally,
given our observations about attitudes towards deaggregation and potential
causes of the diminished response to the CIDR Report, it would be interesting
to explore alternate constructions of a report, such as a report ranked by
deaggregation factor or a report that highlighted recent increases in
deaggregation, that might evoke a more effective response from network
operators.

%- consult with the Internet operations community further

Finally, there are potentially interesting opportunities in considering the
role that CPR governance institutions could play in helping to solve some of
the collective action problems facing the Internet and Internet operators. This
is not so much an extension of this work as suggestions for future directions
based on this work. First, as Ostrom observed in her list of empirical design
patterns for long-standing CPR governance institutions, there is a need for
sanctions to ``back up'' norms and punish opportunism---something which the
CIDR Report lacked. However, in a globally distributed Internet community that
transcends borders and laws, many of the traditional economic or legal tools
that might normally be used to implement sanctions are ineffective or
unavailable. It would thus be interesting to consider technical and social
mechanisms that could possibly be used to implement sanctions or otherwise
disincentivize opportunistic behavior that is against the collective interests
of Internet operators and users. The development of such mechanisms would
likely require a degree of collective action itself, which leads more generally
to the need for appropriate forums for participants to come together to propose
and commit to collective action. While forums such as network operator meetings
and the IETF exist today, these may not be the correct forums, both because of
their existing focus and history, and because of participants fear of the
appearance of collusion and the resulting consequences from antitrust law.
Such work could be a valuable contribution to thinking about improvements to
and altogether new Internet architectures.

% TODO in terms of mechanism and complexity, are CPR institutions + sanctions
% more efficient than markets for things that aren't necessary truly valuable
% (like routing table slots?)



% Economic arguments and palatability in considering replacement protocols in
% the IETF

% Could a "cartel" of DFZ/Tier-1 providers start this because they have the
% disciplinary ability? We can look to the minimum prefix length requirements in
% the IPv4 days (more research required) and in IPv6 with the /32 minimum with
% Verizon, and how this got washed away by an interest in serving the market --
% apparently the incentive to curtail routing table growth just isn't there...?

% While the norms and values embedded in the Internet community may have been
% helpful in past, it's not clear that these are or will continue to hold (cite
% Tussle in Cyberspace) and so we probably shouldn't expect the CIDR report to
% work in the same way, especially with expansion into other regions where the
% culture/Internet ethos isn't the same, or where people aren't aware of the
% history, or loosely/not-at-all connected to the community that does still
% exist.

% As ostrom points out, there's a meta-problem of supplying instutions and
% collective action in bringing about the instiutions one wants to exist to
% solve the first-order collective action problem in the first place.

% Citadinni et al. also point out that the bad guys have stayed the same
% generally, since 2001 -- again, not helpful
%
% This suggests that automated sanctions or pricing might be more effective.
%
% The challenge of agreeing to supply an institution that does anything more than
% provide information (i.e. sanctions) multilaterally is potentially problematic
% in looking beyond the CIDR Report.