[Gzz-commits] gzz/Documentation/misc/hemppah-progradu mastert...

[Hermanni Hyytiälä]

CVSROOT:        /cvsroot/gzz
Module name:    gzz
Changes by:     Hermanni Hyytiälä <address@hidden>      02/11/28 03:17:15

Modified files:
        Documentation/misc/hemppah-progradu: masterthesis.tex 
                                             progradu.bib 

Log message:
        Refs

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/gzz/gzz/Documentation/misc/hemppah-progradu/masterthesis.tex.diff?tr1=1.10&tr2=1.11&r1=text&r2=text
http://savannah.gnu.org/cgi-bin/viewcvs/gzz/gzz/Documentation/misc/hemppah-progradu/progradu.bib.diff?tr1=1.24&tr2=1.25&r1=text&r2=text

Patches:
Index: gzz/Documentation/misc/hemppah-progradu/masterthesis.tex
diff -u gzz/Documentation/misc/hemppah-progradu/masterthesis.tex:1.10 
gzz/Documentation/misc/hemppah-progradu/masterthesis.tex:1.11
--- gzz/Documentation/misc/hemppah-progradu/masterthesis.tex:1.10       Wed Nov 
27 10:19:31 2002
+++ gzz/Documentation/misc/hemppah-progradu/masterthesis.tex    Thu Nov 28 
03:17:15 2002
@@ -200,7 +200,6 @@
 
 \subsection{Coral}
 
-
 \section{Hybrid architecture}
 
 \subsection{Yappers}
@@ -230,6 +229,10 @@
 
 \section{Scalability}
 
+\section{Interopeability}
+
+\section{Performance}
+
 \section{General issues related to resource discovery}
 
 \subsection{Expressiveness}
@@ -258,7 +261,7 @@
 
 \subsection{Anonymity}
 
-\chapter{Overview of Gzz system}
+\chapter{Gzz system}
 
 \section{Overview of Gzz}
 
@@ -276,539 +279,7 @@
 
 \section{Benefits}
 
-\chapter{Conclusion}
-
-
-- - - -
-
-(http://cubicmetercrystal.com/alpine/discovery.html)
-Please notice this not my own text. I'll use this text as supporting text.
-Later, I'll write my own text and add correct refs !!!
-
--hemppah
-- - - -
-
-Current discovery methods are not suitable for large decentralized networks
-Current centralized methods of discovery that are acceptable for dedicated 
servers 
-hosting relatively static content break down when applied to large peer based 
networks. 
-Current decentralized methods lack the efficiency, flexibility and performance 
to be 
-effective in large networks.
-
-Searching the internet and other large networks is currently a very 
centralized process. 
-All of the major search engines such as Google rely on very large databases 
and servers to 
-process queries. These servers and storage systems are very expensive to build 
and maintain, 
-and often have problems keeping information they contain current and relevant.
-
-Search engines are also limited as to the sites they can crawl to obtain the 
data stored in 
-their databases. Your typical peer based network client is far beyond their 
grasp. This 
-makes the vast amount of data available within each peer unknown via this 
traditional 
-method. Data stored in databases accessed via HTML forms and CGI queries is 
also outside 
-the reach of traditional web crawlers.
-
-Peer based networks, such as Freenet and Gnutella rely on a different approach 
to searching. 
-In some cases this is a shared index or external indexing system. In other 
cases this may 
-entail querying specific peers or groups of peers until the resource is 
located (or you grow 
-tired of the search).
-
-All of these approaches lack the flexibility and performance for use in large 
peer based networks.
-
-
-
-Resource discovery in peer based networks is critical to the value of the 
network as a whole
-
-The main benefit provided by peer based networks is the fact that they allow 
access to all 
-kinds of information and resources which were previously unavailable. This may 
be files and 
-documents of interest, or computing power for complex computational tasks.
-
-An important feature of these decentralized peer networks is that their 
perceived value is 
-directly related to the quantity and quality of the resources available within 
them. More 
-resources can be added by increasing the number of peers within the network. 
Thus, the value 
-of the network grows as its popularity increases, which further increases its 
growth, etc, etc.
-
-There comes a point, however, at which more peers no longer increase the 
number of resources 
-available to each peer, and may even cause availability of resources to drop. 
If the network 
-cannot locate resources within the large numbers of peers, or locating 
resources becomes 
-exponentially more expensive as the size of the network grows, it will be 
forever crippled at 
-this threshold.
-
-The ability to locate resources efficiently and effectively regardless of 
network size is 
-therefore critical to the value and utility of the network as a whole.
-
-
-
-Locating resources requires a diverse amount information to be widely 
effective.
-
-Effective discovery methods must rely on a large variety of information about 
the desired 
-resources, typically in the form of metadata.
-
-Metadata varies widely between each kind of resource described. This data can 
be as simple 
-as a filename and SHA-1 hash value, or as detailed as a full cast and credits 
roster for a 
-motion picture. How this meta data is interpreted can also vary widely between 
types of resources. 
-A search for a given amount of processor time for a complex/grid computation 
may require checking 
-system resources, such as scheduled jobs and system load before a reply can be 
provided.
-
-Metadata can vastly improve the accuracy and efficiency of a search, which 
directly affects the 
-utility and popularity of the network.
-
-Support for a wide variety of meta data and searching options is critical to 
the value and 
-utility of any peer based network.
-
-
-
-Any discovery mechanism for large peer based networks must provide a minimum 
set of features
-
-To summarize, an effective discovery mechanism is critical to the value and 
utility of a peer 
-based network. To be effective a discovery mechanism must support a minimum of 
features including:
-
-      Efficient operation in small or large networks
-      Efficient operation for small or large numbers of resources
-      Support a wide variety of meta data and query processing
-      Provide accurate, relevant information for each query
-      Resistant to malicious attack or exploitation
-
-Existing Decentralized Discovery Methods
-
-A short description and assessment of existing decentralized discovery 
mechanisms is provided to 
-compare with a new approach presented in this document.
-
-
-
-All existing discovery methods fail to meet all the desired requirements for 
use in large networks.
-
-There are a number of existing decentralized discovery methods in use today 
which use a variety 
-of designs and architectures. All of these methods have various strengths 
which make them attractive 
-for certain circumstances, however, none of them meet all the criteria desired 
for use in large 
-peer based networks.
-
-The major types of discovery methods we will examine are:
-
-      Flooding broadcast of queries
-      Selective forwarding/routing of queries
-      Decentralized hash table networks
-      Centralized indexes and repositories
-      Distributed indexes and repositories
-      Relevance driven network crawlers
-
-
-
-
-
-Flooding broadcast systems do not scale well
-
-The original Gnutella implementation is a prime example of a flooding 
broadcast discovery mechanism. 
-This type of method has the advantage of flexibility in the processing of 
queries. Each peer can 
-determine how it will process the query and respond accordingly. Unfortunately 
this type of method 
-is efficient only for small networks.
-
-Due to the broadcast nature of each query, the bandwidth required for each 
query grows exponentially 
-with a linear increase in the number of peers. Rising popularity will cause 
the network to quickly 
-reach a bandwidth saturation point. This causes fragmentation of the network 
into smaller groups of 
-peers, and consumes a large amount of bandwidth while in operation.
-
-Segmentation of the network reduces the number of peers visible and the 
quantity of resources 
-available. Queries must be sent over and over again to try and compensate for 
the reduced range of 
-queries in a highly segmented network. It may take a large amount of time for 
a suitable number of 
-peers to be queried, which further reduces the effectiveness of this approach.
-
-This type of discovery mechanism is very susceptible to malicious activity. 
Rogue peers can send out 
-large numbers bogus queries which produce a significant load on the network 
and disproportionately 
-reduce network effectiveness.
-
-False replies to queries can be formulated for spam / advertising purposes, 
which reduces the accuracy
- of the queries.
-
-
-
-Selective forwarding systems are susceptible to malicious activity
-
-Selective forwarding systems are much more scalable than flooding broadcast 
networks. Instead of 
-sending a query to all peers, it is selectively forwarded to specific peers 
who are considered likely 
-to be able to locate the resource. While this approach greatly reduces 
bandwidth limitations to 
-scalabality, it still suffers from a number of shortcomings.
-
-First and foremost is susceptibility to malicious activity. Due to the fact 
that a much smaller 
-number of peers receive the query, it is vastly more important that each of 
these peers be reputable 
-for this operation to be effective.
-
-A rogue peer can insert itself into the network at various points and misroute 
queries, or discard 
-them altogether. Results can be falsified to degrade the accuracy and 
relevance of results. Depending 
-on the pervasiveness and operation of this peer(s), performance can be 
degraded significantly.
-
-Any system that relies on trust in an open, decentralized network will 
inevitably run into problems 
-from misuse and malicious activity.
-
-Each peer must also contain some amount of additional information used to 
route or direct queries 
-received. For small networks this overhead is negligible, however, in larger 
networks this overhead 
-may grow to levels that are unsupportable.
-
-While an improvement over flooding broadcast techniques, this approach is 
still not suitable for a 
-large peer based network.
-
-
-
-Decentralized hash table networks do not support robust search
-
-Decentralized hash table networks further optimize the ability to locate a 
given piece of information. 
-Every document or file stored within the system is given a unique ID, 
typically an SHA-1 hash of its 
-contents, which is used to identify and locate a resource. The network and 
peers are designed in such a 
-way that a given key can be located very quickly despite network size. This 
type of system does have 
-severe drawbacks which preclude its use as a robust searching and discovery 
method.
-
-Since data is identified solely by ID, it is impossible to perform a fuzzy or 
keyword search within 
-the network. Everything must be retrieved or inserted using an ID.
-
-These systems are also susceptible to malicious activity by rouge peers. A 
rogue peer may misdirect 
-queries, insert large amounts of frivolous data to clutter the keyspace, or 
flood the network with 
-queries to degrade performance. In such hierarchial or shared index systems 
these attacks can inflict 
-much more damage than the bandwidth and CPU resources required to initiate 
them. 
-
-(Amplifying effect on the attack)
-
-While more resilient than flooding broadcast networks, and efficient at 
locating known pieces of 
-information, these networks are still not able to perform robust discovery in 
large peer based networks.
-
-
-
-Centralized indexes are expensive and legally troublesome
-
-Centralized indexes have provided the best performance for resource discovery 
to date. However, they 
-still entail a number of significant drawbacks which preclude their use in 
large peer based networks.
-
-The most serious issue is cost. The bandwidth and hardware required to support 
large networks of peers 
-is prohibitively expensive. Scaling this kind of network requires substantial 
capital investment and may 
-still reach limits that unsupportable.
-
-Recent court rulings cast serious doubt about the liability involved in using 
centralized servers to 
-index resources in a peer based network. It has been said that the recent 
legal precedents require any 
-such system to monitor usage and activity of the network exactly to ensure 
that no types of copyright 
-violations are occurring. The ability to monitor and enforce this requirement 
is quite challenging, and 
-may be too much of a risk.
-
-Centralized index systems are not suitable solutions for resource discovery in 
large peer based networks.
-
-
-
-Distributed indexes are dificult to maintain and susceptible to malicious 
activity
-
-Distributed indexes eliminate the need for expensive centralized servers by 
sharing the indexing burden 
-among peers in the network. Legal vulnerability is greatly decreased by 
removing central control of indexing 
-operations. When designed correctly, these types of networks provide the best 
performance and scalability 
-of any solution. Even more so than most centralized solutions.
-
-The most difficult problem with these types of indexing systems is cache 
coherence of all the indexed data. 
-Peer networks are much more volatile, in terms of peers joining and leaving 
the network, as well as the 
-resources contained within the index. The overhead in keeping everything up to 
date and efficiently distributed 
-is a major detriment to scalability.
-
-There have been a number of proposals and implementations of shared index 
systems which address this problem. 
-Unfortunately distributed indexes encounter problems in the following 
situations:
-
-      The number of peers supporting the index network is large
-      Many peers join and depart the network maintaining the index
-      The amount of data to be indexed is significant
-      The meta data for the indexed data is very diverse
-      Malicious peers exploit the trust implicit in a shared index
-
-All large peer based networks exhibit these features, making a distributed 
index system incredibly 
-complicated.
-
-Their susceptability to malicious attack is also increased. Rogue peers may 
insert large amounts 
-of frivolous data which burdens the shared index as well as reducing the 
accuracy of searches 
-within it. There is a much larger degree of trust placed on each peer, due to 
that fact that 
-each peer must handle and search the indexed data correctly, and also that 
each peer help maintain 
-(in terms of bandwidth and physical storage) the shared index equally (or at 
least to the best of 
-their ability given finite resources) This makes resilience in the face of 
rogue peers extremely difficult.
-
-Supporting a wide range of meta data can also be difficult. An XML schema may 
be provided to 
-contain this data, however, tracking the meta data in addition to keys or 
names significantly 
-increases the indexing overhead, further reducing scalability of the network. 
Since each peer 
-must search its section of the index at given times, each peer must also be 
able to understand 
-the meta data as it relates to the query it is processing. This is also a 
significant burden, 
-as diverse peers may or may not understand the meta data and how to interpret 
it.
-
-Distributed indexing systems as they currently exist cannot provide robust 
discovery in large 
-networks. I hope that will change at some point in the future, as this would 
be the best solution hands down.
-
-
-
-Relevance driven network crawlers lack support for proactive queries and 
diverse data
-
-Relevance driven network crawlers are a different approach to the resource 
discovery problem. Instead 
-of performing a specific query based on peer request, they use a database of 
existing information the 
-peer has accumulated to determine which resources it encounters may or may not 
be relevant or interesting 
-to the peer.
-
-Over time a large amount of information is accrued which is analyzed to 
determine what common elements 
-the peer has found relevant. The crawler then traverses the network, usually 
consisting on HTML documents 
-for new information which matches the profile distilled from previous peer 
information.
-
-The problem with this system is that it lacks support for proactive queries 
for specific information, as 
-it is directed by past information. Support for a wide variety of resources is 
also missing, since the 
-relevance engine expects a certain kind of data on which it can operate. This 
usually consists of HTML 
-or or other text documents.
-
-Finally, this type of discovery can be too slow for most uses. The time 
required for the crawler to 
-traverse a significant amount of content can be prohibitively long for uses on 
modems or DSL connections.
-
-Relevance driven network crawlers are not suitable for discovery in large 
networks.
-
-
-Optimizations to Existing Discovery Methods
-
-Many of the afore mentioned discovery methods have been tweaked and tuned in 
various ways to increase the 
-efficiency and accuracy of their operation. A few of this enhancements are 
described below.
-
-
-
-Intelligence and hierarchy in flooding broadcast networks
-
-The Gnutella network has come a long way since its conception in April of 
2000. The first new feature is 
-increased intelligence in the peers in the network. The second is the use of 
hierarchy to differentiate high 
-bandwidth, dedicated peers from slower, less powerful peer clients.
-
-The original Gnutella specification was very simple and intended for small 
groups of peers. This simple protocol
- lacked the forethought required for scaling in larger networks. Once the 
network gained popularity it became 
- obvious to all involved that additional features were required to avoid the 
congestion in a larger, busy network.
-
-One popular modification was denying access to gnutella resources to web based 
gnutella clients. These web 
-interfaces allowed a large number of users to search the network without 
participating, and thus placed a 
-large load on the network with no return value. Many clients will no longer 
share files with peers who themselves 
-do not share.
-
-Other expensive protocol operations, such as unnecessary broadcast replies 
were quickly replaced with 
-intelligent forwarding to intended destinations.
-
-Connection profiles were implemented to favor higher bandwidth connections 
over slower modem connections so 
-that slow users were pushed to the outer edges of the network, and no longer 
presented a bottle neck to network 
-communication.
-
-Expanding on this theme, the Clip2 Reflector was introduced to allow high 
bandwidth broadband users to act as 
-proxies for slower modem users.
-
-All in all the Gnutella network and related systems have made vast progress. 
In many cases they may provide 
-adequate performance despite their intrinsic weakenesses.
-
-
-
-Catalogs and meta indexes in distributed hash table networks
-
-The desire to allow flexible keyword and meta data searching in distributed 
hash table networks has resulted in 
-various methods to catalog the data contained within them.
-
-A new project called Espra stores catalog documents within Freenet itself that 
describe the resources represented 
-by their hash key identifier. Additions and searching can be performed on 
these catalogs to locate resources 
-efficiently and quickly within the network.
-
-Other networks consist of similar methods which keep the catalog or index in 
external web servers or documents.
-
-The main drawback with this approach is that it requires the maintenance of 
these catalogs. Locating a given catalog 
-or index in the first place may also be a problem.
-
-These methods have provided a much needed ability to search for resources in 
these distributed hash table networks, 
-however, they still lack the robustness and flexibility desired in an optimal 
solution.
-
-
-
-Keyword search for distributed hash table networks
-
-Another use of distributed hash tables is keyword searching using individual 
hash values for each keyword in a query. 
-Each keyword produces a set of matches, which can then be combined for complex 
muti-word keyword searches.
-
-This approach looks very promising, as it retains the attractive performance 
and scalability of distributed hash tables 
-while providing the flexiblity of keyword / metadata based searching. There 
should be some implementations of this 
-coming out sometime in 2002, however, none are in a stable, useable state as 
of this time.
-
-Implementations of searching over distributed hash tables need to solve two 
hard problems. The first is support for 
-load distribution of hotspots: very popular hash keys. Some keywords are very 
popular and these keywords could drive 
-an unsupportable amount of traffic to a single node (or small set of nodes) in 
the distributed hash table network. 
-There must be some mechanism for many nodes to share the load of popular 
keywords.
-
-The second problem is the protection of the insert mechanism in the keyword 
indexes. It is hard to ensure that all 
-users returning hits for a given keyword are legitimate, and false or 
malicious results stored/appended at a given 
-keyword could severely impact the performance of the search.
-
-Once these problems are solved or minimized searching over distributed hash 
table networks could provide a very robust 
-search mechanism for large peer networks.
-
-
-
-Hybrid networks using super peers and self organization
-
-A popular type of hybrid network has been implemented by FastTrack and used in 
the Morpheus and KaZaa media sharing 
-applications. This approach has also been implemented in the now defunct Clip2 
Reflector, and the JXTA Search implementation.
-
-This type of network replaces the dedicated central servers used in indexing 
content with a large number of super peers. 
-These peers have above average bandwidth and processing power which allows 
them to take on this additional workload without 
-affecting performance a great deal. Every peer in the network contacts one or 
more of these super nodes to search for 
-matches to a given query.
-
-Super peers are selected automatically based on some kind of bandwidth and 
memory/cpu metric. Often there is some kind of 
-colloboration between super peers to relay queries if no matches are found 
locally, and to provide super peer nodes to new clients.
-
-This architecture provides the best solution to date. By avoiding fully 
centralized servers these networks have been a bit 
-more resiliant legally (although KaZaa and FastTrack are currently in legal 
manuevers).
-
-These types of networks appear to be the current sweet spot for searching 
networks. Napster was too centralized, and gnutella 
-not enough. Meeting at the middle with a hybrid super peer network gives you 
the best of both worlds.
-
-There are still a number of problems with this architecture. Despite being 
less of a legal target than a true centralized 
-server, they are still 'mini' centralized servers in function. Given the 
recent court rulings these nodes would have to monitor 
- and filter content to avoid possible copyright infringement violations. 
Requiring each node to contain a list of all filter 
- information would be near impossible to implement given the current size of 
filters used by the RIAA alone. The now defunct 
- OpenNap server network was a distributed collection of smaller centralized 
servers, and they were threatened out of existence. 
- It is likely that once the encryption used in FastTrack has been circumvented 
that the super peers would be a prime target for RIAA/MPAA nasty grams.
-
-Support for robust meta data information is also difficult to provide with 
this type of architecture. This requires each super 
-node to support all of the meta data types used in matching queries for the 
resources it indexes. For a wide variety of meta 
-data this would require a large amount of overhead in synchronizing support 
for this meta data in all super nodes as well as 
-adding the functionality for specific meta data types in each super node.
-
-These super nodes are also prime targets for malicious attack. Since each peer 
they are connected to provides them with index 
-information, as well as queries, it takes a small amount of effort for a peer 
to send a large volume of false index information 
-as well as large numbers of bogus queries. Depending on the specific 
implementation of these super peers this may cause 
-excessive memory usage, truncated indexes, and low performance.
-
-Finally, this type of network relies the on the generosity of peers in the 
network to provide these super peers. In current 
-implementations this is an optional feature and may or may not be feasible in 
a large network.
-
-
-
-
-
-An Adaptive Social Discovery Mechanism for Large Peer Based Networks
-
-We now describe the architecture of an adaptive social discovery mechanism 
that is designed to work efficiently, 
-effectively, and in a scalable manner for large peer based networks.
-
-
-
-Social discovery implies a direct, continued interaction between peers in the 
network
-
-One of the fundamental differences with this approach is that it requires a 
direct connection between each peer and the 
-peers it communicates with. We will see that this impacts a large number of 
the requirements for a robust discovery mechanism.
-
-Each peer directly controls which peers it communicates with, how bandwidth is 
consumed, and how the network is used. 
-This provides powerful abilities to resist abuse of the network, allocate 
bandwidth according to the users preferences, 
-and last but not least, allows many optimizations of the discovery process 
which would not be available otherwise.
-
-Each connection is also much longer lived than a typical TCP connection. These 
connections can be re-established when a 
-dialup user changes IP addresses or a NAT user changes ports. They persist as 
long as the peers agree to communicate.
-
-This longevity of connections allows peers to maintain a history of their 
interaction with each of their peers which in 
-turn is used for reputation management and optimization of discovery 
operations within the network.
-
-
-
-Simple, low overhead messaging forms the foundation of peer communication
-
-At the base of this discovery implementation is the use of UDP for simple, low 
overhead messaging via small data packets. All 
-communication between peers is performed through a single UDP socket. An 
application level multiplexing protocol supports the 
-large number of direct connections with very little overhead. This is similar 
to the way that TCP and UDP connections are 
-multiplexed over IP using port numbers.
-
-All discovery operations require a certain amount of communication between 
peers to locate a given resource. In large 
-decentralized networks this often consumes the majority of bandwidth 
available. By making the messaging protocol as compact 
-and lightweight as possible, we reduce the overhead required for sending any 
given message.
-
-
-
-Connection persistence allows profile and performance tracking of peers
-
-The base protocol also uses much longer connection lifetimes between peers. 
Connections can be re-established if the 
-application is restarted, if the modem line disconnects, and if the ports 
change on a NAT firewall. As long as the peers 
-wish to remain connected they may do so.
-
-The reason for this feature is to maintain a history for each peer. This 
history is used to build a profile of the peer 
-to determine how 'valuable' it is for discovery operations, and how many 
resources it has used.
-
-Peers that are outright malicious can be identified by providing no value, yet 
using large amounts of bandwidth or other 
-resources. Their connection is then terminated.
-
-Peers who consume but do not share resources will in turn be viewed as very 
low quality peers and their connections terminated 
-as well. This prevents abuse of the network, or the tragedy of the commons 
effect, and encourages peers to provide resources 
-and be good neighbors.
-
-
-
-Past query responses are used to optimize resource discovery
-
-The actual search for resources within the network is accomplished by sending 
a single compact query packet to each 
-peer in the group to be queried. This proceeds in a linear fashion until a 
sufficient number of resources are 
-located, or the user terminates the query.
-
-This would be a rather slow and inefficient operation if no further 
optimizations were made. To increase the 
-efficiency of the discovery operation the profile associated with each peer is 
used to determine the order in 
-which each peer is sent a query packet.
-
-Peers who have responded with relevant, quality resources in the past will 
have a higher quality value in their 
-profile than those peers who have not.
-
-By querying the peers with the higher quality value first, the chances of 
finding a resource quickly are greatly 
-increased. This in turn decreases the total amount of bandwidth and time 
required for a search.
-
-
-
-Social discovery and profiling encourages sharing and good behavior
-
-Most searching networks provide little incentive for peers to provide more 
resources. The 'Free Loaders' problem 
-has been stated quite often when discussions about peer networking arise. 
There have been some attempts to 
-eliminate free loading and bad behavior using agorics or reputation, however, 
these methods have proven very difficult to apply.
-
-In a social discovery network each peer must contribute or risk loosing the 
peers that it is connected to. Likewise, 
-if you want to be able to connect to high quality peers, you must strive to be 
a high quality peer yourself. This 
-is all handled autonomously given the adaptive nature of peer organization 
during queries and other operations.
-
-As peers continually refine their peer groups, the bad or low quality peers 
will be dropped and replaced with new 
-peers who might have better characteristics. In this way, good behavior and 
large numbers of quality resources are 
-rewarded and encouraged.
-
-
-
-Distinct groups of peers are supported for distinct types of discovery
-
-In many cases a user will search for various types of resources on the same 
network. While a peer may be a very 
-good peer for one type of query, it may be very poor for another. For this 
reason groups of peers are supported 
-so that peers can be queried when most appropriate.
-
-This prevents high quality peers from getting poor ratings during queries 
which they do not support, and allows 
-increased efficiency for the discovery operation by providing groups of peers 
tuned to the specific type of 
-discovery operation.
-
-For example, one set of peers may be used to locate classical recordings, 
while another may be used to locate small 
-animation files. Each peer may be useful for one type of query and not the 
other, and groups ensure that peers are 
-treated appropriately based on their performance for specific types of queries.
-
-
-
-Extensions are supported for a wide range of meta data and functionality
-
-Another core feature of this approach is the use of modular extensions to the 
discovery operations and application 
-functionality. A protocol extension ID is specified within each query packet. 
Any third party can define a set of 
-meta data or protocol extensions and assign it a unique extension ID. Any 
client which supports that extension can 
-now process the meta data appropriately for much greater flexibility and 
accuracy during the discovery operation.
-
-Often there is additional processing required for a given set of protocol or 
meta data extensions. This is supported 
-using dynamic modules which contain the required code to process this 
information. These modules can be loaded and 
-unloaded at runtime according to a users needs.
-
-This modular, extensible system provides the flexibility to support a wide 
range of meta data and protocol extensions 
-to further increase the quality and value of responses received.
-
-
-
-Adaptive social discovery relates directly to the interaction of a user with 
his/her peers
-
-Taken as a whole, this process maps closely to the actual interaction that 
occurs between a user and the peers 
-(s)he communicates with in the network.
-
-Groups of peers with similar interests will organize spontaneously as they 
would in the physical world, and can 
-remain in continued interaction with each other as long as they find the 
relationship valuable.
-
-Conversely, those peers which do not contribute to the group or attempt to 
attack the peers outright will find 
-themselves ostracized until they cease their undesirable behavior.
-
-By taking advantage of this style of interaction the quality, performance and 
flexibility required for decentralized 
-resource discovery in large peer based networks can be implemented 
successfully. 
+\chapter{Conclusion} 
 
 
 \bibliographystyle{gradu}
Index: gzz/Documentation/misc/hemppah-progradu/progradu.bib
diff -u gzz/Documentation/misc/hemppah-progradu/progradu.bib:1.24 
gzz/Documentation/misc/hemppah-progradu/progradu.bib:1.25
--- gzz/Documentation/misc/hemppah-progradu/progradu.bib:1.24   Tue Nov 26 
09:12:37 2002
+++ gzz/Documentation/misc/hemppah-progradu/progradu.bib        Thu Nov 28 
03:17:15 2002
@@ -1095,3 +1095,47 @@
        howpublished = {http://www.xdegrees.com/}
 }
 
+
+%Search improvments in p2p networks
address@hidden,
+       author = "S. Rhea and J. Kubiatowicz",
+       title = "Probabilistic location and routing",
+       booktitle = "In Proceedings of INFOCOM 2002",
+       year = "2002",
+       url = 
"http://oceanstore.cs.berkeley.edu/publications/papers/pdf/rk-infocom-2002.pdf";
+}
+
+%Tangler publishing system
address@hidden,
+       author = {Marc Waldman and David Mazi},
+       title = {Tangler: a censorship-resistant publishing system based on 
document entanglements},
+       booktitle = {Proceedings of the 8th ACM conference on Computer and 
Communications Security},
+       year = {2001},
+       isbn = {1-58113-385-5},
+       pages = {126--135},
+       location = {Philadelphia, PA, USA},
+       doi = {http://doi.acm.org/10.1145/501983.502002},
+       publisher = {ACM Press},
+}
+
+%Security considerations for structured peer-to-peer overlay networks
address@hidden, 
+       title = "Security for structured peer-to-peer overlay networks", 
+       author = {M. Castro, P. Druschel, A. Ganesh, A. Rowstron, and D. 
Wallach},
+       booktitle = {Fifth Symposium on Operating Systems Design and 
Implementation (OSDI'02)},
+       location = {Boston, MA},
+       month = {December},
+       year = {2002}
+}
+
+%Tarzan anonymizing network layer (security)
address@hidden:ccs9,
+  title = {Tarzan: A Peer-to-Peer Anonymizing Network Layer},
+  author = {Michael J. Freedman and Robert Morris},
+  booktitle = {Proceedings of the 9th {ACM} Conference on Computer and
+Communications Security ({CCS-9})},
+  year = {2002},
+  month = {November},
+  address = {Washington, D.C.},
+}
+