A TAXONOMY OF PROGRAMMABLE HTTP PROXIES FOR

ADVANCED EDGE SERVICES

Delﬁna Malandrino

Dipartimento di Informatica ed Applicazioni “R.M. Capocelli”

Universit

a di Salerno

Via S. Allende, Baronissi (SA), 84081, Italy

Vittorio Scarano

Dipartimento di Informatica ed Applicazioni “R.M. Capocelli”

Universit

a di Salerno

Via S. Allende, Baronissi (SA), 84081, Italy

Keywords:

HTTP proxy, programmable proxy, edge services.

Abstract:

In this paper, we present the state of the art in the ﬁeld of programmability in HTTP proxies. In particular, we

ﬁrst deal with programmability and show how it is a crucial requirement to easily realize and assemble edge

services that can enhance the quality and the user perception of the navigation into a crowded and confusing

World Wide Web. Then, we compare some of the most used HTTP proxies to provide an analysis of their

programmability and, ﬁnally, show some evidence of successful edge services realized on top of existing

programmable HTTP proxy frameworks.

1 INTRODUCTION

The architecture of the World Wide Web has under-

gone a signiﬁcant change in the last 10 years. The

simple client-server architecture, that was one of the

key factors of the World Wide Web success, has been

integrated with multiple interconnected nodes dis-

seminated across the Internet so that end-users can ex-

perience advanced services with better performances

(low latency), fault tolerance and high availability.

Despite the success of these architectures, the in-

troduction of new services into existing networks is

often an expensive and time-consuming process. In

order to improve and simplify this process, crucial has

become the requirement of programmability in all the

Internet infrastructures. A programmable infrastruc-

ture offers a quick prototyping and assembling of new

services as well as easy and responsive modiﬁcations

of existing ones.

In this research area, the interest is mainly devoted

to frameworks that make possible an easy develop-

ment of new services as well as enhancing the ex-

isting core functionalities of the distributed environ-

ments that provide such services.

Proxy servers are intermediary entities that can be

placed anywhere along an information stream in the

content path between origin server and client systems

aiming to improve the performance of the WWW and

its scalability.

Intermediaries can be developed in order to pro-

vide intelligent services beyond simple and tradi-

tional caching and content replication services. These

value-added services, also called edge services, in-

clude content adaptation and personalization, local-

ization, content aggregation, mobility and ubiquitous

access to the Internet content.

The edge of the network is a point that resides in

the path between clients and servers where content

processing is requested. Many are the potential places

where edges can store and change content or provide

services, in particular (for our purposes) examples are

proxy servers.

The goal of the paper is to show how proxy

servers represent the ideal place for adding program-

mability functionalities. For example, intermediaries

could choose to handle the request immediately, exe-

cute content processing to produce and deliver value-

added content, access resources on remote locations

and so on. By splitting the issuing of the requests

from making and managing them, it is possible to in-

crease the ﬂexibility (newfunctionalities can be added

without changing the existing ones) and the power of

the Web.

Organization of the paper: In the following section,

231

Malandrino D. and Scarano V. (2005).

A TAXONOMY OF PROGRAMMABLE HTTP PROXIES FOR ADVANCED EDGE SERVICES.

In Proceedings of the First International Conference on Web Information Systems and Technologies, pages 231-238

DOI: 10.5220/0001233802310238

 SciTePress

we brieﬂy described how simple proxy services have

evolved into complex edge services. Then we de-

tail, in Section 3, what we mean by a programmable

infrastructure. In Section 4 we describe how each

programmable proxy infrastructure is rated against

the dimensions we outlined before. Finally, in Sec-

tion 5 we describe some advanced edge services that

can be obtained only by leveraging on highly conﬁg-

urable and programmable proxy infrastructures and

conclude the paper in Section 6 with some ﬁnal re-

marks.

2 FROM PROXY TO EDGE

SERVICES

As the nature of the services delivered by the Web

grew more and more rich and heterogeneous, it be-

came necessary to design and implement more and

more elaborate caching strategies and infrastructures

that later evolved into the “Content Delivery Net-

works” (CDNs) (B. Krishnamurthy, 2001) that, ﬁrst,

introduced the concept of “Edge” of the network.

While content networks create a virtual overlay of

Edge intermediary servers, on the top of the packet

network, for an efﬁciently delivering of customized

content to the end users, the Edge Services Overlay

Network moves one step further, by providing a ser-

vices infrastructure that involve the development of

another intelligent overlay network around CDNs.

The idea of the Edge overlay networks is to de-

velop and deploy software entities, strategically dis-

tributed through the network, to provide content adap-

tation and other complex functionalities (personaliza-

tion, location-aware data insertion, etc) on the data

ﬂow exchanged between clients and servers. These

software entities must be able to provide a value-

added on the content delivered to end users and, in

the meantime, ensure robustness, high availability and

scalability performance.

Intermediary software infrastructures exhibit fea-

tures that match the above requirements. They realize

a transparent deployment by appearing as a client for

existing servers and as a server for existing clients,

reduce requirements on clients and complexity on

servers and, ﬁnally, they are able to provide com-

plex intermediary functionalities, without involving

software/hardware modiﬁcations both on client and

server systems. In addition, it is well-know that prox-

ies are essential to improve Internet access by reduc-

ing user-perceived latency and assure a better quality

of services, to which WWW users are most interested

in.

Intermediary systems, layered on existing network

infrastructures, can be easily programmed and de-

ployed to operate as an overlay network of connected

or cooperative components in order to efﬁciently pro-

vide advanced edge services.

Many are the application ﬁelds of edge computing

services that our vision encompasses: from the geo-

graphical personalization of the navigation of pages,

with insertion/emphasis of content that can be related

to user geographical location, to translation services

(IBM Almaden Research Center, 1999); from sup-

port for group navigation and awareness (Barrett and

Maglio, 1998; Calabr

o et al., 2003) for social navi-

gation (Barra et al., 2002) to advanced services for

bandwidth optimization such as adaptive compression

and format transcoding (Ardon et al., 2003a; Hori

et al., 2000; WebSphere, ). A particularly useful ap-

plication ﬁeld for edge services is the accessibility

of Web sites. These applications are usually heavy-

weight and therefore, assuring their scalability is par-

ticularly critical. A typical example is an ubiqui-

tous service that provides text-to-speech (Barra et al.,

2003) translation of the pages navigated by the user,

regardless of the location and conﬁguration of the lo-

cal machine (i.e. without installing particular soft-

ware).

Of course “traditional” proxy services like caching

can be further enhanced by combination with ad-

vanced edge services: an example can be offering

the user a “search engine” that operates only on

cached/personalized/translated pages.

3 PROGRAMMABILITY OF

PROXIES

In this section we introduce a detailed taxonomy of

characteristics that intermediary systems should ex-

hibit for quick development and easy deployment of

advanced edge computing services. In particular we

identify and investigate the following important fea-

tures.

• Programmability.

With this term we mean the requirement of changing

or updating the functionalities provided by a software

infrastructure as well as the possibility to entirely de-

velop new intermediary services. The core function-

alities must be initially provided and off-loaded into

the intermediary infrastructure and, after program-

mers must develop and deploy new application ser-

vices.

Programming under existing intermediaries cannot

involve the same power, efﬁciency, and generality like

developing an open system from scratch. This means

that we need of an execution environment, in which

a compositional framework should provide the ba-

sic components for developing new services, and a

programming model to make this execution environ-

ment a programmable environment. Such program-

WEBIST 2005 - WEB INTERFACES AND APPLICATIONS

232

ming model should provide APIs, software libraries

and languages (object-oriented or others), for pro-

gramming and deploying new services into the inter-

mediary infrastructure.

Moreover, by enhancing the distinction between

the core network infrastructure and the programming

model for developing new services, programmers can

develop services without taking care of details con-

nected to scalability, fault tolerance and availability.

• Conﬁgurability.

With this term we mean the requirement of conﬁgur-

ing the set of services that can be invoked during Web

transactions. In particular we can distinguish into two

types of possible conﬁgurations: (a) a simple conﬁg-

uration for adding, removing, enabling or disabling

functionalities and (b) a more detailed conﬁguration

in which service parameters can be speciﬁed by ask-

ing the user to ﬁll-out Web-based forms (for example

by providing a downgrade quality image for an im-

age transcoding service). In this context another im-

portant requirement concerns the possibility to load,

install and execute new services at run-time without

recompiling and restarting the running software in-

frastructure.

Finally service parameters could be speciﬁed

through a Graphical User Interface (GUI) to simplify

the process of services conﬁguration. Moreover, that

GUI could to appear useful for monitoring the behav-

ior of the system (invoked services, HTTP requests

and responses, etc).

• Adaptability.

With this term we mean the need to address the is-

sues of the existing and emerging wireless and mobile

technologies that require content adaptation and dy-

namic reconﬁguration based on the speciﬁc context:

limited devices capabilities (i.e. small screen sizes)

limited or intermittent connection to the wired net-

work, etc.

• Centralized or Distributed System Architecture.

The architecture of a software intermediary architec-

ture can be expressed by components that can be in

execution on a single host (centralized architecture)

or, rather, by components that are distributed on dif-

ferent hosts and that collaborate to provide Internet

services and other complex functionalities such as

load balancing, fault tolerance, high availability and

performance scalability.

• Lifecycle support of applications.

This requirement is needed to support the deploy-

ment and un-deployment of proxy services, by mak-

ing these tasks automatic and accessible by both lo-

cally and remote locations. Application deployment

is an important system functionality that allows client

to anytime-anywhere access to provided services and

applications. By making the deployment an automatic

task (i.e. wizard) it is possible to add new functional-

ities into the intermediary system without taking care

of the complexity of the system itself, and without

manipulating and changing system conﬁguration ﬁles

(which is a typical error-prone task).

Active deployment allow to add new services into

the network or to customize existing ones to math new

application demands without restarting or reconﬁgur-

ing Web proxy systems that host the services.

• Personalization.

User Proﬁling represents the an efﬁciently way,

within mobile and ubiquitous environments, to pro-

vide personalized services according to the more and

more different and heterogeneous client devices that

have access to the more and more dynamic and in-

teractive Internet content. Users proﬁles are also im-

portant to make difference between groups of users.

If the intermediary system supports this functionality,

some authentication mechanism must be provided.

• Services activation.

An important feature that a distributed service in-

frastructure could exhibit is the mechanism for ser-

vices activation. A rule-based services activation can

be useful to dynamically determining the data path

through the various components (that provide the ser-

vices) allowing arbitrarily complex boolean expres-

sions instead of simple URL and content-type match-

ing. Such rules should be dynamically modiﬁed to

add or remove services, or to change their conditions.

• Maintainability of software.

Finally, an important feature concerns the availabil-

ity (source code) of the developed software in order

to ensure a swift adaptation of the software to the

demands for new advanced services. In fact, open-

source software is considered the most efﬁcient so-

lution when system infrastructure has to be easily ex-

tended and enhanced with new system functionalities.

4 INTERMEDIARY SYSTEMS

In this section we brieﬂy describe the interemediary

systems that are, then, compared in Table 1.

RabbIT. RabbIT (RabbIT, ) is a Web HTTP proxy

that accelerates the delivery of Web contents to end

users by compressing text pages and images, by re-

moving unnecessary parts of HTML pages (back-

ground images, advertisements, banners, etc.) and ﬁ-

nally by caching ﬁltered documents before forward-

ing them to the clients.

RabbIT proxy, written in Java, has a modular archi-

tecture that allow an easy deﬁnition and implementa-

tion of new functionalities. In addition, these modules

can be conﬁgured at runtime (no parameter conﬁgu-

ration is provided), but we need to restart the proxy to

keep all chosen settings.

A TAXONOMY OF PROGRAMMABLE HTTP PROXIES FOR ADVANCED EDGE SERVICES

233

RabbIT Web proxy is characterized by a central-

ized structure, it is programmable, partially conﬁg-

urable (no services parameters are allowed), it does

not provide support for adaptability, user proﬁling and

automatic deployment. New services (ﬁlters) can be

added to the proxy but we need to access conﬁgura-

tion ﬁles and manually specify new settings.

Mufﬁn. Mufﬁn (Mufﬁn, 2004) is a Web HTTP

proxy that provides functionalities to remove cookies,

kill GIF animations, remove advertisements, add, re-

move, or modify any HTML tag, remove Java applets

and Javascript code, etc.

It is programmable and conﬁgurable proxy, new

services or ﬁlters can be developed through a set of

provided APIs and can be added at runtime, using the

provided graphical interface, without restarting the

proxy.

WebCleaner. WebCleaner (Webcleaner, ) is a ﬁlter-

ing HTTP proxy that provides functionalities for re-

moving advertisements, banners, ﬂash and javascript

code, for reducing image (by size) and compressing

HTML pages (with gzip), etc.

WebCleaner proxy is programmable and highly

conﬁgurable, but it does not provide support for user

proﬁling. New services can be added into the system

through a graphical user interface. Through this GUI

new Filter Modules can be added by specifying ap-

propriate conﬁguration parameters.

FilterProxy. FilterProxy (FilterProxy, 2002) is a

modular HTTP proxy that provides functionalities for

compressing HTML pages, image transcoding, etc. It

is programmable and highly conﬁgurable, new ser-

vices (or ﬁlters) can be dynamically added to the sys-

tem, and their conﬁguration is realized through web-

based forms, or editing a conﬁguration ﬁle. Services

can be chained in order to provide complex function-

alities. FilterProxy is written in perl, and is quite fast.

Privoxy Web Proxy. Privoxy (Privoxy, 2004) is a

Web proxy with advanced ﬁltering capabilities for

protecting privacy, modifying Web page content, con-

trolling access, and removing advertisements, ban-

ners, pop-ups, etc.

Privoxy has a very ﬂexible conﬁguration and can

be customized according to individual user needs.

Speciﬁcally, Privoxy can be conﬁgured with various

conﬁguration ﬁles. By typing a speciﬁc URL into

the Web browser it is possible to select a set of ac-

tions and Privoxy’s conﬁguration parameters can also

be viewed at the same page.

It provides a set of API and a detailed documenta-

tion on how to develop and deploy new services.

WBI. Web Based Intermediaries (WBI) is a dy-

namic and programmable framework, developed at

IBM Almaden Research Center (IBM Almaden Re-

search Center, 1999), whose main goal is to personal-

ize the Web by realizing an architecture that simplify

the develop of intermediary applications. WBI de-

ﬁnes a programming model that can be used to imple-

ment all form of intermediaries, from simple server

functions to complex distributed applications. WBI is

freely usable but its sources are not publicly available.

A WBI transaction is deﬁned as a complete HTTP

request-response and ﬂows through a combination of

the four type of basic stages, called MEGs: RequestE-

ditor and Editor MEGs, that modify HTTP requests

and responses respectively, the Generator MEG that

given an HTTP request produce the corresponding re-

sponse and ﬁnally, the Monitor MEG that monitors

HTTP requests and response without modify them at

all. A group of such MEGs is called WBI plugin.

WBI provides a rule-based system for dynamically

determining the data path through the various MEGs

allowing complex boolean expressions. With WBI,

rules can be dynamically altered to add or delete

MEGs, or to change their ﬁring condition. Finally

the WBI GUI provides a convenient way to manage

and administer WBI, and to help users to debug the

plugins loaded into the proxy.

AT&T Mobile Network iMobile. iMobile (Rao

et al., 2001) is a proxy-based platform designed to

provide personalized mobile services. It provides a

modular architecture that support accesses from vari-

ous mobile devices to various information spaces.

The main component of iMobile platform is iProxy

(AT&T Labs-Research, 2002), a programmable proxy

server that provides an environment for develop-

ing personalized services, which are implemented as

building blocks in Java.

The modularity into the iMobile framework is

achieved through three different abstractions: (a) de-

vlets, that support the provision of iMobile services

to different type of mobile devices, (v) infolets, that

provide a communication mechanism between iMo-

bile servers and the information sources, and (c) the

applets, that encapsulate the services functionalities.

The iMobile project was subsequently extended

into e new project, iMobile EE (Chen et al., 2003), to

address the requirements of security, high availability

and performance scalability.

eRACE. The extensible Retrieval Annotation

and Caching Engine (eRACE) (Dikaiakos and

Zeinalipour-Yiazti, 2001a) is a modular, program-

mable and distributed intermediary infrastructure

whose main goal is to provide personalized services

for a wide range of client devices (Desktop PC,

WEBIST 2005 - WEB INTERFACES AND APPLICATIONS

234

mobile and thin clients), such as, personalization,

customization, ﬁltering, aggregation, transformation

both on wireline and wireless Internet. The client

requests are scheduled for execution by an eRACE

scheduler which can implement different policies

for load balancing. Authentication and proﬁling are

managed by a Service Manager integrated into the

system.

The most important component of the system

is WebRACE (Dikaiakos and Zeinalipour-Yiazti,

2001b) a proxy system that deals with WWW in-

formation. It consists of a Web crawler, a ﬁltering

processor and an object cache. These components

can be distributed on several nodes and communicate

through socket links.

TACC. The BARWAN project (Katz et al., 1996)

by UC/Berkeley has the goal to provide intermediary

systems to support ubiquitous access to Internet ser-

vices from mobile and thin clients. An important sys-

tem component of this project is the programmable

proxy architecture, TACC, that acts as intermediary

between servers and mobile clients.

The main component of the TACC compositional

framework is the worker, that encapsulates the service

functionality. Workers, that represent the building

blocks of TACC applications, can be chained in or-

der to provide more complex functionalities, such as

transformation, aggregation, caching and customiza-

tion of Web content (Brewer et al., 1998; Fox et al.,

1998a). Other modules handle user proﬁling and

provide mechanisms for load balancing and fault-

tolerance.

ALAN. ALAN, acronym of Application Level Ac-

tive Network (Fry and Ghosh, 1999), is a network ar-

chitecture that provides mechanisms to dynamically

load code, or proxylet, to facilitates Web content de-

livery (MacLarty and Fry, 2001). Communications is

enhanced by Dynamic Proxy Server that are located at

strategic points along the content path between client

and server.

Proxylet can be downloaded, as JAR archive,

from HTTP servers into a Dynamic Proxy Server

infrastructure, making their deployment very easy

(URL to reference proxylets). The main contribu-

tions of this approach are: (a) the fast deployment

of new communication services on demand without

the drawbacks of the Active Networks (deployment

of new elements at network routers) and (b) a plat-

form that provide ﬂexible and value-added services

(WWW streaming audio, WWW compression, etc).

WebPADS. WebPADS (Chuang and Chan, 2005),

acronym of Web Proxy for Actively Deployable Ser-

vices, provides a framework that facilitates the devel-

opment of add-on services or mobilets, which can be

actively deployed and migrated across Web proxies,

according to the dynamic changes in wireless envi-

ronments.

WebPADS, developed in Java, provides a set of

APIs for developing mobilets and mobile appli-

cations. Prototype implementations, to prove the

framework’s beneﬁts, include compression and image

transcoding services.

Finally, mobilets can be chained together in a

nested order to allow service aggregation and an easy

reconﬁguration if changes in the mobile environments

occur.

MARCH. MARCH (Ardon et al., 2003b) is a dis-

tributed content adaptation architecture that provides

functionalities for adapting Web content according to

the capabilities of client devices that access to Internet

services.

While MARCH enables the creation of an over-

lay network of intermediary entities, deployed be-

tween clients and servers, it adopts a server-centric

approach, by delegating, for example, the decision

of which proxy invoke for a given service (front end

component of the system) to a centralized mobile

aware server.

An important feature of this system architecture is

that it allows the dynamic composition of services

(proxy chain) for a given set of operating conditions

(CPU, memory, etc). MARCH is implemented in Java

and prototype implementations include examples of

transcoding services and a compression service.

Squid. Among the “classical” (i.e. non program-

mable) proxy, particularly popular is SQUID (Squid

Proxy, ). It is a high-performance proxy caching

server for web clients, supporting FTP, gopher, and

HTTP data objects. Unlike traditional caching soft-

ware, Squid handles all requests in a single, non-

blocking, I/O-driven process. Squid acts as an agent,

accepting requests from clients (such as browsers)

and passes them to the appropriate Internet server. It

stores a copy of the returned data in an on-disk cache.

The real beneﬁt of Squid emerges when the same data

is requested multiple times, since a copy of the on-

disk data is returned to the client, speeding up Internet

access and saving bandwidth.

Despite its efﬁciency in dealing as proxy cache,

Squid is not explicitly mentioned here in the compar-

ison, since it lacks a real support for programmabil-

ity. In fact its main functionalities include caching

of HTTP, FTP and other URLs, transparent caching,

proxy-ing for SSL, HTTP server acceleration. Most

of SQUID-based products focus on caching, ﬁrewall

technologies, security, network monitoring and con-

tent ﬁltering (Cerberian, ).

A TAXONOMY OF PROGRAMMABLE HTTP PROXIES FOR ADVANCED EDGE SERVICES

235

Table 1: Proxy Comparison. Notice that RabbIT is only partially conﬁgurable, i.e. it does not allow to specify parameters

for services.

Programmability Conﬁgurability Adaptability Centralized / Lifecycle Personalization Activation Maintainability

Distributed Support Services

RabbIT

√

∼ - C - - -

√

Mufﬁn

√ √

- C

√

- -

√

Privoxy

√ √

- C - - -

√

WebCleaner

√ √

- C

√

- -

√

FilterProxy

√ √

- C

√

- -

√

WBI

√ √

- C

√

Free

iMobile

√ √ √

eRACE

√ √ √

TACC

√ √ √

ALAN

√

- - D

√ √ √

Free

WebPADS

√ √ √

√

MARCH

√ √ √

5 ADVANCED EDGE SERVICES

The current trend in the ﬁeld of pervasive computing

is how to enable universal access to the Internet con-

tent, that is, how to address the critical issue of de-

veloping technologies that are able to allow the trans-

formation, or content adaptation, of Web content for

adapting it to the different and heterogeneous client

devices that exhibit different capabilities in terms of

network connectivity, processing power, storage, dis-

play capabilities, etc.

More precisely, content adaptation or transcod-

ing represents the process of converting Web content

from one form to another, and it is typically carried

out by content servers or, on-the-ﬂy, by proxy servers

at the Edge of the network.

Content adaptation refers to two different abstrac-

tions, that is, personalization, the adaptation of the

Web content according to user preferences, locations

and contexts, and transcoding, the process of of tailor-

ing Web content to the capabilities of the client device

and the network connections.

Adaptation can take place at client-side, server-

side or at intermediary entities in between. In a

proxy-based approach, content adaptation can be per-

formed on-the-ﬂy, without overload the clients (as in

client-based approaches) and without allow multiple

variants of the same content (as in server-side ap-

proaches).

Several proposal, academic (Fox et al., 1998a; Fox

et al., 1998b; Fox and Brewer, 1996; Zenel, 1999)

and commercial products (WebSphereEdgeServices,

), are HTTP proxy-based, since they explore proxy

systems to adapt multimedia Internet content.

Scone (Weinreich et al., 2001) is a modular Java

framework whose main gaol is to improve the navi-

gation and orientation on the Web, by generating new

views of Web documents, offering workgroup tools

to support collaborative navigation, enriching Web

pages with new navigational elements, etc.

Extraction of useful and relevant part of Web doc-

uments has many applications, such as displaying for

devices with limited size capabilities, speech render-

ing for users with visual disabilities, summarizing of

Web contents, etc.

The process of extracting part of a document re-

quires two important steps: (a) the parsing of HTML

Web pages (to remove unnecessary elements, links,

images, etc), and (b) the building of the DOM repre-

sentation of such pages (for their reformatting). These

operation can be efﬁciently realized on the top of in-

termediary systems (Gupta et al., 2004) in order to

reduce complexity both on client and server systems.

Content aggregation refers to the generation of con-

tent at intermediary systems. More precisely, an ag-

gregator queries one or more Web sites for speciﬁc

content (news, search. etc) and collects and formats

the results for the presentation to the clients. In the

TOP GUN Wingman proxy (Fox et al., 1998b), the

aggregation service allows the user to enter some in-

formation describing what is being requested. The re-

quest is processed by a service module into the proxy

and the response is formatted by an HTML proces-

sor. This aggregator works (and produce results) for

Yahoo, HotBot, AltaVista, DejaNews, Yahoos stock

quotes service, and TripQuest.

Due to its popularity and its ease of use, the Web

is an attractive platform to support distributed coop-

erative works (Cabri et al., 1999). These systems

can help users to quickly ﬁnd information among

WEBIST 2005 - WEB INTERFACES AND APPLICATIONS

236

the growing amount of information available on the

WWW.

A proxy-based architecture can be employed to

support synchronous cooperation on the Web. By

programming a proxy-framework, many cooperative

functionalities can be provided to allow users within

a common workgroup to share and cooperate toward

a common goal.

Proxy-based frameworks can be also programmed

to operate as recommender systems, assisting users in

their navigation and making users aware of others’s

information activities, interests and knowledge. The

WebMemex system (Macedo et al., 2003) was devel-

oped on top of a high level architecture, an augmented

Web proxy server, that provides capture, linking, user

authentication, storage, retrieval and access capabili-

ties.

Another interesting functionality that can be of-

fered by proxy systems concerns the recording of

Web navigation. An example is WebMacros (Safonov

et al., 2001) a proxy-based system for automating

repetitive user interactions with the Web by recording

and replaying user navigation.

Finally, intermediary systems can be used as visu-

alization systems (Hong and Landay, 2001), in or-

der to help Web design teams to run usability tests

and analyze the collected data, and Web users for re-

visitation of Web pages by allowing dynamic drawing

of animated graphs of the user’s paths as the surf the

Web.

6 CONCLUSION

The examples shown in the last section witness the

usefulness of programmability in intermediary in-

frastructures. Without a full support toward the devel-

opment of new services, advanced mechanisms can-

not be easily realized, deployed and managed, while

quick prototyping and manageability represent crucial

requirements to assure that programmers can quickly

respond to mutations of data format, content or stan-

dards that are so common in the Web, nowadays. Our

comparison shows how this ﬁeld is now proposing

several platforms that share a common objective, to

ensure that advanced edge services can be efﬁciently

and effectively realized.

An important direction toward standardization is

being conducted by the IETF Working Group for

“Open Pluggable Edge Services” (OPES)

. Their goal

is to deﬁne an open standard that allows interme-

diaries to provide services on the data ﬂow. In

See (Barbir et al., 2003) for the description of the

architecture and the Web site http://standards.

nortelnetworks.com/opes for the overall activities

of this WG.

OPES, intermediaries can also employ local or remote

servers (called “callout servers”) in order to facilitate

the efﬁcient delivery of complex services. OPES rule-

set are applied in order to choose which service to

apply to the data ﬂow. Communication between dis-

patchers and callout servers occurs through a proto-

col that must obey the requirements for OPES Callout

Protocols speciﬁed in (Beck et al., 2002).

Of course, as OPES gains momentum, program-

mability of proxies will get an additional character-

istics to be measured against, namely the compatibil-

ity with an open standard to interconnect components.

Therefore, it is clearly foreseeable that, if OPES gets

enough support, the environments for programmable

proxies will take into account their openness to stan-

dards.

REFERENCES

Ardon, S., Gunningberg, P., LandFeldt, B., Y. Ismailov,

M. P., and Seneviratne, A. (2003a). MARCH: a dis-

tributed content adaptation architecture. International

Journal of Communication Systems, Special Issue:

Wireless Access to the Global Internet: Mobile Radio

Networks and Satellite Systems., 16(1).

Ardon, S., Gunningberg, P., LandFeldt, B., Y. Ismailov,

M. P., and Seneviratne, A. (2003b). March: a dis-

tributed content adaptation architecture. International

Journal of Communication Systems, Special Issue:

Wireless Access to the Global Internet: Mobile Radio

Networks and Satellite Systems., 16(1).

AT&T Labs-Research (2002). iProxy: a Programmable

Proxy. http://www.research.att.com/sw/tools/iproxy/.

B. Krishnamurthy, C. Wills, Y. Z. (2001). On the use perfor-

mance of contend distribution network. In Proceed-

ings of SIGCOMM IMW.

Barbir, A., Chen, R., Hofmann, M., H.Orman, and Penno,

R. (2003). An Architecture for Open Pluggable

Edge services (OPES). http://www.ietf.org/internet-

drafts/draft-ietf-opes-architecture-04.txt.

Barra, M., Grieco, R., Malandrino, D., Negro, A., and

Scarano, V. (May 2003). TextToSpeech: an heavy-

weight Edge computing Service. In Poster Proc. of

International World Wide Web Conference. ACM

Press.

Barra, M., Maglio, P., Negro, A., and Scarano, V. (2002).

GAS: Group Adaptive System. In Proceedings of In-

ternational Conference on Adaptive Hypermedia and

Adaptive Web-based Systems (AH 2002). ACM Press.

Barrett, R. and Maglio, P. P. (1998). Adaptive Communities

and Web Places. In Proceedings of 2

Workshop on

Adaptive Hypertext and Hypermedia, HYPERTEXT

98., Pittsburgh (USA). ACM Press.

Beck, A., Hofmann, M., Orman, H., Penno, R., and Terzis,

A. (2002). Requirements for OPES Callout Protocols

An Architecture for Open Pluggable Edge Services

A TAXONOMY OF PROGRAMMABLE HTTP PROXIES FOR ADVANCED EDGE SERVICES

237

(OPES). http://www.faqs.org/ftp/internet-drafts/draft-

ietf-opes-protocol-reqs-02.txt.

Brewer, E., Katz, R., Amir, E., Balakrishnan, H., Chawathe,

Y., Fox, A., Gribble, S., Hode, T., Nguyen, G., Pad-

manabhan, V., Stemm, M., Seshan, S., and Hender-

son., T. (1998). A Network Architecture for Hetero-

geneous Mobile Computing. In IEEE Personal Com-

munication Magazine, 5(5):8–24.

Cabri, G., Leonardi, L., and Zambonelli, F. (1999). A

proxy-based framework to support synchronous coop-

eration on the web. Softw. Pract. Exper., 29(14):1241–

1263.

Calabr

o, M. G., Malandrino, D., and Scarano, V. (2003).

Group Recording of Web Navigation. In Proceedings

of the HYPERTEXT’03. ACM Press.

Cerberian. Cerberian web ﬁltering software.

http://www.cerberian.com/main.html.

Chen, Y.-F., Huang, H., Jana, R., Jim, T., Hiltunen, M.,

John, S., Jora, S., Muthumanickam, R., and Wei, B.

(2003). iMobile EE: an enterprise mobile service plat-

form. Wirel. Netw., 9(4):283–297.

Chuang, S. N. and Chan, A. T. (2005). Active service for

mobile middleware. WWW: Internet and Web Infor-

mation Systems Journal.

Dikaiakos, M. and Zeinalipour-Yiazti, D. (2001a). A dis-

tributed middleware infrastructure for personalized

services. Technical Report TR-2001-4, University of

Cyprus.

Dikaiakos, M. and Zeinalipour-Yiazti, D. (2001b). We-

bRACE: A Distributed WWW Retrieval, Annotation,

and Caching Engine. In Proceedings of PADDA01:

International Workshop on Performance-oriented Ap-

plication Development for Distributed Architectures.

FilterProxy (2002). FilterProxy.

http://ﬁlterproxy.sourceforge.net/.

Fox, A. and Brewer, E. A. (1996). Reducing WWW latency

and bandwidth requirements by real-time distillation.

In Proceedings of the 5

International World-Wide

Web Conference. ACM Press.

Fox, A., Chawathe, Y., and Brewer, E. A. (1998a). Adapting

to Network and Client variation using active proxies:

Lessons and perspectives. IEEE Personal Communi-

cations, 5(4):10–19.

Fox, A., Goldberg, I., Gribble, S., Lee, D., Polito, A., and

Brewer, E. (September 1998b). Experience with top

gun wingman, a proxy-based graphical web browser

for the usr palmpilot. In Proceedings of the IFIP In-

ternational Conference on Distributed Systems Plat-

forms and Open Distributed Processing (Middleware

’98).

Fry, M. and Ghosh, A. (1999). Application level active net-

working. Comput. Networks, 31(7):655–667.

Gupta, S., Kaiser, G. E., Grimm, P., Chiang, M. F., and

Starren, J. (2004). Autmatic content extraction of html

Documents. World Wide Web Journal.

Hong, J. I. and Landay, J. A. (2001). Webquilt: A frame-

work for capturing and visualizing the web experi-

ence. In Proceedings of the 10

International World

Wide Web Conference, Hong Kong. ACM Press.

Hori, M., Kondoh, G., Ono, K., Hirose, S., and Singhal, S.

(2000). Annotation-Based Web Content Transcoding.

In Proc. of the 9

Intl. World Wide Web Conf., Ams-

terdam (The Netherland). ACM Press.

IBM Almaden Research Center (1999).

Web Based Intermediaries (WBI).

http://www.almaden.ibm.com/cs/wbi/.

Katz, R. H., Brewer, E. A., Amir, E., Balakrishnan, H., Fox,

A., Gribble, S., Hodes, T., Jiang, D., Nguyen, G. T.,

Padmanabhan, V., and Stemm, M. (1996). The Bay

Area Research Wireless Access Network (BARWAN).

In Proceedings of the 41st IEEE International Com-

puter Conference, page 15. IEEE Computer Society.

Macedo, A. A., Truong, K. N., Camacho-Guerrero, J. A.,

and da GraÇa Pimentel, M. (2003). Automat-

ically sharing web experiences through a hyperdoc-

ument recommender system. In HYPERTEXT ’03:

Proc. of the 14th ACM Conf. on Hypertext and Hy-

permedia, pages 48–56. ACM Press.

MacLarty, G. and Fry, M. (2001). Policy-based content de-

livery: an active network approach. Computer Com-

munications, 24(2):241–248.

Mufﬁn (2004). Mufﬁn Proxy. http://mufﬁn.doit.org/.

Privoxy (2004). Privoxy Web Proxy.

http://www.privoxy.org/.

RabbIT. RabbIT proxy. http://rabbit-

proxy.sourceforge.net/.

Rao, H. C.-H., Chang, D.-F., Chen, Y.-F., and Chen, M.-F.

(2001). iMobile: a proxy-based platform for mobile

services. In Wireless Mobile Internet, pages 3–10.

Safonov, A., Konstan, J. A., and Carlis, J. V. (May 2001).

Webmacros - a proxy-based system for automating

user interactions with the web. In Poster Proc. of

International World Wide Web Conference. ACM

Press.

Squid Proxy. Squid Proxy. http://www.squid-cache.org/.

Webcleaner. Webcleaner Filter Proxy.

http://webcleaner.sourceforge.net/.

WebSphere. IBM Websphere Transcoding Publisher.

http://www-3.ibm.com/software/webservers/

transcoding.

WebSphereEdgeServices. Websphere Edge Server.

http://www-3.ibm.com/software/webservers/

edgeserver.

Weinreich, H., Obendorf, H., and Lamersdorf, W. (2001).

The look of the link, concepts for the user interface of

extended hyperlinks. In Proceedings of the 12th ACM

Conference on Hypertext and Hypermedia, University

of Aarhus, Denmark, pages 19–28. The Association

for Computing Machinery, ACM New York.

Zenel, B. (1999). A general purpose proxy ﬁltering mecha-

nism applied to the mobile environment. Wirel. Netw.,

5(5):391–409.

WEBIST 2005 - WEB INTERFACES AND APPLICATIONS

238