=Paper=
{{Paper
|id=Vol-552/paper-2
|storemode=property
|title=Knowledge Federation: Necessity and Required Technologies
|pdfUrl=https://ceur-ws.org/Vol-552/Tanaka-KF08.pdf
|volume=Vol-552
}}
==Knowledge Federation: Necessity and Required Technologies==
https://ceur-ws.org/Vol-552/Tanaka-KF08.pdf
Knowledge Federation: Necessity and Required
Technologies
Yuzuru Tanaka
Meme Media Laboratory, Hokkaido University
N13 W8, Sapporo 0608628, Japan
{tanaka}@meme.hokudai.ac.jp
Abstract. This paper focuses on knowledge federation from different
aspects; its taxonomy, its necessity, and its required technologies. There
are three different aspects of knowledge federation, namely, semantic
federation, service federation, and view federation. Each knowledge fed-
eration uses at least one of these aspects. The latter half of this paper
focuses on ad hoc knowledge federation of Web resources including Web
applications. Ad hoc knowledge federation deals with two of the three as-
pects, i.e., service federation and view federation. This paper reviews the
current author’s R&D on meme media technologies and ad hoc knowl-
edge federation based on them, and then proposes an extension of the
Web toward ‘the memetic Web,’ in which users’ collaborative reediting
and redistribution of Web resources accelerates the evolution of extended
Web resources. Finally, it introduces a new version of the meme media
system IntelligentPad as an enabling technology for the memetic Web.
This new version runs on de facto standard browsers empowered by Sil-
verlight plug-in, and requires no IntelligentPad kernel running on clients.
1 Introduction
The Internet has been and still is continuously changing our lives. It is pervading
the economy and the private sphere, thus boosting the pace at which modern
information and communication technologies are penetrating everyone’s life, ev-
ery science and engineering discipline, and almost every business. Despite the
fact that billions of Web resources are available world-wide and no slowdown of
the Web’s growth is in sight, users typically make little use of the richness of the
Internet.
One well-known problem is the difficulty of finding what a user is interested
in, due to a lack of a suitable schema and a query language to store and retrieve
resources. A second problem is the difficulty of making mutually related Web
resources work together, due to a lack of suitable interfaces or conversely a
plethora of interfaces, formats, and standards. Here in this paper, let me call the
first problem ‘the findability problem,’ and the second ‘the reusability problem.’
There is a similar prominent problem in practically every large enterprise,
in the larger healthcare institutions and in literally every governmental agency:
enterprise application integration. The key problem is that legacy systems have
been introduced over many decades. Each department has no means to know
what kinds of applications are used and how they are related with each other in
another department. Most of these systems have never been designed to work
together, and the potential for synergetic effects dwindles away. Both the find-
ability and reusability problems exist there. Such situations can be commonly
observed even in those enterprises which, in 70s and 80s, made large investments
to fully integrate their data and applications by introducing database manage-
ment systems. Such paradoxical situations were brought by rapid and ad hoc
introduction of many different types of client applications, and also by the needs
for the interoperation with Web resources.
The diversity, heterogeneity, autonomy, and openness of both resources and
their usages are the inherent characteristics of current information system envi-
ronments. These characteristics are inherently out-of-control, and incompatible
with conventional top-down system integration methods.
It seems that human creativity will always lead to newly designed systems
and solutions that go beyond earlier standards and overall pictures of whole
systems, and are thus not fully compatible with the established and widely used
systems. Creating and updating standards in information and communication
technologies is of course needed, but there is little hope that this alone could
ever solve the enterprise application integration problem, or alleviate the above-
mentioned problems that face individuals, enterprises and communities when
dealing with the diversity of the Web.
The ‘findability problem’ is partially solved either by providing indices for
useful resources, or by semantically restructuring resource fragments. While the
first approach has developed Web search engines using resource indices, the sec-
ond approach has introduced portal sites, Semantic Web technologies [1], Topic
Map tools [2], Web ontologies [3], ontology mediation technologies [4], and se-
mantic inference technologies [5]. The semantic restructuring is further classified
into two different views of semantic entities, i.e., as universals and as names.
The first view assumes the existence of a single universal semantic structure
among entities. This first view worked well for such compiled sets of resources as
databases, but not for such an ever-evolving open set of resources like the Web.
The latter requires semantic restructuring based on the second view, which may
independently restructure mutually overlapping resource and resource-fragment
sets in different semantic schemes, and later use ontology mediators to consis-
tently merge these semantic schemes into a virtually integrated scheme.
The ‘reusability problem’ addresses how to extract some fragments you want
to reuse from the resources we find, how to customize each of them for fitting
it to its reuse context, and how to coordinate them work together to perform a
required job. Here the coordination means a huge variety of different goals, which
include the construction of a single portal Web page showing these fragments,
the application of some of them as functions to the others as data sets, the
combination of their functions to compose a new function, and the composition
of a new visual Web resource embedding the visual representations of these
fragments.
Semantic restructuring of resources and resource fragments, and functional
and/or visual combination of resource fragments give a partial solution respec-
tively to the findability problem and to the reusability problem. Both of these
topics share the same characteristics; the diversity, heterogeneity, autonomy,
network-distribution, and openness of resources and their usages, the virtual
or real extraction of useful resource fragments into structured formats, and the
definition of interrelation and/or interoperation among them to treat them as
a single virtual complex information and/or functional entity. We use the word
‘knowledge federation’ to denote such a single virtual entity defined from Web
resources, or such a defining action. The word federation emphasizes the goal to
form a virtual unity without restricting any of the diversity, heterogeneity, au-
tonomy, network-distribution, and openness of both resources and their usages.
This paper focuses on knowledge federation from different aspects; its tax-
onomy, its necessity, and its required technologies. First, it points out that there
are three different factors of knowledge federation, namely, semantic federation,
service federation, and view federation. Then it focuses on ad hoc knowledge
federation of Web resources including Web applications. Ad hoc knowledge fed-
eration deals with two of the three factors, i.e., service federation and view
federation. It reviews the current author’s R&D on ad hoc knowledge federation,
and proposes an extension of the Web toward ‘the memetic Web,’ in which users’
collaborative reediting and redistribution of Web resources accelerates the evo-
lution of extended Web resources. Finally, this paper introduces our new version
of the IntelligentPad system as an enabling technology for the memetic Web.
This new version runs on de facto standard browsers empowered by Silverlight
plug-in, and requires no IntelligentPad kernel running on clients.
2 Taxonomy of Knowledge Federation
The word ‘knowledge’ in computer science was initially used in a very limited
meaning in 70s and 80s to denote what can be described in some ‘knowledge rep-
resentation language’ for the problem solving by machines. The same word in
dairy conversation denotes facts, information and skills acquired through experi-
ence or education, theoretical or practical understanding of a subject, language,
etc., or awareness or familiarity gained by experience of a fact or situation. Today,
we use computers to deal with all these different types of knowledge by repre-
senting them as some kinds of objects such as documents, tools and services. IO
relations and/or functions of applications and services can be also considered as
knowledge fragments.
The Web is expanding its capability to deal with all these different types of
knowledge for their sharing in a society and/or in an enterprise, and to promote
their reuse in different contexts for different purposes. Recently, some philoso-
phers such as Davis Baird are trying to expand the definition of knowledge to
denote not only linguistically expressible knowledge, but also ‘thing knowledge’,
i.e., knowledge embodied as some artifact object [6]. Computerization of the
design and manufacturing of artifacts is also accumulating their thing knowl-
edge as reusable and sharable knowledge. Only some Web documents satisfy the
definition of linguistically expressible knowledge, while all the Web resources
including documents, tools, and services are artifacts, and satisfy the definition
of thing knowledge. Our definition of ‘knowledge’ in this paper includes both
linguistically expressible knowledge and thing knowledge.
Knowledge resources denote sharable and reusable different types of such
knowledge represented on computers. The Web basically represents knowledge
resources as compound documents with embedded nonfunctional multimedia
contents and/or functional contents such as application tools and services. The
compound document architecture was initially proposed around mid 1980s as
an extension of multimedia documents. While a multimedia document embeds
in itself multimedia contents, a compound document looks like a multimedia
document and can embed in itself not only multimedia contents but also any
interactive visual objects with some functionality. Web documents as well as
Web applications are compound documents. Compound documents treated in
the same system are based on the same compound document architecture.
Knowledge resources require some media to externalize them as interactive
visual objects, and to publish them for their reuse by other people. Knowledge
media denote such media since they externalize knowledge resources [7]. Knowl-
edge resources treated in the same system exploit the same knowledge media ar-
chitecture. Compound document architectures are examples of knowledge media
architectures. While compound documents are two dimensional visual represen-
tations of knowledge resources, some knowledge media architectures represent
knowledge resources as three dimensional visual objects or spaces.
The Web provides both a standard knowledge media architecture to pub-
lish knowledge resources as Web documents, and an open repository of Web
documents for publishing, sharing, and reusing them among different people
distributed over the world. Web documents are compound documents.
The word ‘federation’ was probably first introduced to IT areas in mid 80s by
Dennis Heimbigner in the context of a federated database architecture [8], and
then secondarily in late 90s, by Bill Joy in a different context, namely, federation
of services [9]. In a federated database architecture, a collection of independent
database systems are united into a loosely coupled federation in order to share
and exchange information. The term federation in this context refers to the
collection of constituent databases participating in a federated database. In the
context of service federation, a federation is a set of services that can work
together to perform a task.
In a federated database, a federation consists of components and a single
federal dictionary [8]. The components represent different databases, individual
users, applications, workstations, or other components in an office information
system. The federal dictionary is a specialized component that maintains the
topology of the federation and oversees the entry of new components. Each
component in the federation controls its interactions with other components by
means of an export schema and an import schema. The export schema specifies
the information that a component will share with other components, while the
import schema specifies the nonlocal information that a component wishes to
manipulate. The federated architecture provides mechanisms for sharing data,
for sharing transactions via message types to combine information from several
components, and for coordinating activities among autonomous components via
negotiation.
The idea of federating databases can be extended to federation of Web infor-
mation resources, which we call ‘semantic federation.’ Each constituent database
schema corresponds to the articulation of useful information fragments in some
set of Web information resources, and the semantic structuring of these articu-
lated fragments. Semantic Web and Topic Map technologies are used for such
articulation and semantic structuring. Instead of using export and import schema
to consistently merge different but partially overlapping database schemata, fed-
eration of Web information resources combines more than one semantic structure
over sets of Web information resources to provide a federated single semantic
view of the relationships among those resource fragments.
Federation of services, on the other hand, denotes the definition and execution
of interoperation among services that are accessible either through the Internet or
through peer-to-peer ad hoc communication. It may be classified into two types:
ad hoc federation defined by users and autonomic federation defined by programs.
Ad hoc service federation denotes a technological goal that enables users to define
a service federation without writing a code, while autonomic federation denotes a
service federation in which each service-requesting service autonomously finds a
corresponding service-providing service to form a federation. Most of the service
federation architectures proposed in the past focus on autonomic federation [10–
13]. Only a few of them focus on ad hoc federation of services [14, 15].
Autonomic service federation architectures basically propose two things, i.e.,
a standard communication protocol with a language to use it, and a repository-
and-lookup service that allows each member to register its service-providing
capabilities, and to request a service that matches its demand. For each re-
quest with a specified demand, a repository-and-lookup service searches all
the registered service capabilities for those satisfying the specified demand.
A repository-and-lookup service matches service-requesting queries with corre-
sponding service-providing capabilities. The origin of such an idea can be found
in the original tuple space model Linda [10] and its extension Lime [11] that
copes with mobile objects by providing each of them a dedicated tuple space.
Linda and Lime are languages that use tuples to register service-providing capa-
bilities and to issue service-requesting queries to a repository-and-lookup service
called a tuple space. Java Space [12] and Jini [13] are Java versions of Linda and
Lime architectures.
The idea of federating services can be also extended to federation of knowl-
edge resources over the Web. Knowledge resources include not only Web ser-
vices but also Web applications and Web documents. Federation of knowledge
resources first needs to extract useful fragments from Web resources as services
with or without visual presentations. Especially in case of a Web application
without its API, it needs to extract an IO relation between some input forms
and some output-page fragments as a reusable service [14, 15]. From a Web doc-
ument, it may extract a structured set of data [16]. Then it needs to combine
these data sets and services so that they may work together virtually as a sin-
gle application with or without a composite visual presentation. Federation of
Web resources with extraction and composition of visual presentations needs to
define not only federation of services but also federation of their presentation
views into a single composite presentation view. We call the latter ‘view federa-
tion.’ We may also consider a view federation of Web applications and/or Web
documents without their service federation. For example, some tools help you
extract portlets from different Web resources and to combine them to construct
a portal site without defining any interoperation among these portlets [17].
Knowledge federation is characterized by the following three different aspects;
1. semantic federation,
2. service federation, and
3. view federation.
Each knowledge federation uses at least one of these aspects.
Federation is becoming fundamental to IT. It is the process of combining
multiple technology elements and/or heterogeneous resources into a single vir-
tual entity. Federation is being driven by complexity, and by IT’s need to make
sense of technology systems that have sprawled to unprecedented scale and have
become extremely expensive to maintain and manage. Federation is different
from integration in the following sense. Federation assumes an open networked
environment of heterogeneous, autonomous, and distributed resources, and deals
with open scenarios of information processing. Integration basically targets local
and centralized management and interoperation of resources in a closed environ-
ment, and deals with closed scenarios of information processing.
3 R&D Trends in Knowledge Federation
R&D on semantic federation uses either Semantic Web or Topic Maps as their
basis. As to the extraction of useful Web information in a structured format,
one remarkable effort can be found in DBpedia [18], which is a community effort
to extract structured information from Wikipedia and to make this information
available on the Web. DBpedia provides us with a semantic schema in which
extracted information resources are structured, and allows us to access these
resources through this schema by writing a query in a query language SPAQL.
Soon in a future, we will have lots of such semantically structured Web infor-
mation extracted from various different sets of Web resources. We will need a
semantic federation of them to merge their schemata into a single unified schema
so that we may issue a query through this unified schema.
R&D on service federation of Web services has three mutually overlapping
technology trends. The first trend uses a tuple space or an XML space to federate
Web services [19]. The second trend uses a GRID computing architecture and its
workflow/orchestration tools to federate Web services [20, 21]. The third trend
uses Semantic Web description for semantically federating Web services [22].
This technology is called Semantic Web Service. Semantic Web Service has two
aspects of knowledge federation; service federation and semantic federation.
R&D on service federation of Web resources including Web applications has
two technology trends. One of them is represented by Web mashup tools. The
other is the current author’s knowledge federation framework [14, 15] based on
meme media technologies [23]. Since service federation of Web resources includ-
ing Web applications needs to deal with page presentations of Web applications,
it requires user’s intervention for specifying which portions to extract and how to
arrange them to compose a target Web page. Service federation of Web resources
including Web applications also requires view federation.
Web mashup denotes the user-driven micro-integration of Web data, where
micro integration means data merging, data feeding, data joining, data filtering,
and data annotating of Web data. Web mashup tools range from the feeding
of Web service output data to the API of a specified Web application such as
Google Maps [24] and Google Chart, to advanced mashup editors such as Intel
Mash Maker [25] and Lotus Mashup. Web mashup tools in the first generation
used some specific Web application such as Google Maps as the base application,
and used its API to add, on its page, some graphical objects that represent the
data retrieved from other Web services. The server program of this base Web
application needs to provide such a function through its API. The addition
of graphical objects corresponding to the external data is performed by the
server program of this Web application. Traditional mashup tools are based on
workflow composition of Web services to construct a new Web application or
service. The I/O data transfer in such a composition depends on SOAP, REST
and/or RSS/ATOM feeds. To include a Web application, users must derive a
proxy Web service by using some extraction tool. These mashup tools cannot
deal with Web applications as source resources.
Recently, technologies are moving towards on-screen visual composition of
new resources, including widgets extracted from Web applications. Mash Maker
supports augmentation of Web pages with data from other sources. Users can
define an extractor for the data within an HTML page, and paste registered
widgets onto such pages, connecting them through the relevant extractor. The
widgets are portlets encapsulating Web-application behavior, created by writing
code. Lotus Mashup supports composition within a Web-based workspace. Tools
allow users to create widgets that capture Web application behavior. Widgets
are combined in the workspace. There is a good paper comparing two cultures;
mashing up Web 2.0 and the Semantic Web [26].
In the sequel of this paper, we will focus on ad hoc knowledge federation of
Web resources including Web applications. Ad hoc knowledge federation deals
with both service federation and view federation. This type of knowledge feder-
ation is the one the current author proposed its concepts and architectures in
2004 [27] based on our R&D on 2D and 3D meme media since 1987 [23].
4 Ad hoc Knowledge Federation of Web Resources
Including Web Applications
4.1 Its requirements
The Web is becoming an open repository of a huge variety of knowledge re-
sources. The problem we face today is how to make it easy to select some set
of resources, customize them, and make them work together to meet our ever-
changing demands for new knowledge resources. Commercially available drug
design software systems are a good example of the complex integrated applica-
tion systems available today. Each system provides an integrated environment
comprising modules such as a protein database, a homology search system, a
ligand structure prediction system, and a docking simulator. While such inte-
grated systems are very expensive, for each module you can often find a free Web
service, or a downloadable freeware program. Similar situations are frequently
observed in a range of academic areas. Nonetheless, finding appropriate data
sources and applications from the Web, customizing them, and making them
work together is as difficult and expensive as developing a commercial product
with similar complexity. If there were an easy, generic way for users to edit these
open resources instantaneously to compose the desired complex applications, we
would be able to accelerate substantially the evolution of the knowledge resources
shared by our societies. The existing proposals for federation, such as Jini, pro-
vide no support for instantaneous extraction, customization, or composition of
knowledge resources.
Recently a new research methodology has been developed for the advanced
automation of scientific data acquisition tools, based on collecting data for all
cases without any specific purposes, then later retrieving and analyzing appro-
priate data sets to support a specific hypothesis or conjecture. This new method-
ology is referred to as ‘data-based science’, or ‘data-centric science’. Data-based
science is becoming more and more popular in scientific areas such as genomics,
proteomics, brain science, clinical trials, nuclear physics, astrophysics, material
science, meteorology, and seismology. Data-based science makes data acquisition
independent from data analysis; it leads to large numbers of huge, independent
accumulations of data, and a great variety of data analysis tools. Many such
data accumulations and analysis tools have been made available over the Web,
to encourage their reuse. Advanced utilization of such resources requires both
flexible extraction of appropriate data sets from these diverse sources, and flex-
ible application of appropriate analysis tools.
Such an approach, although currently performed with lots of manual op-
erations and programming, has enabled the comparative analysis of seemingly
unrelated large data sets, leading to the discovery of interesting correlations such
as those between the El Niño-Southern Oscillation and the weather in Europe,
volcanic eruptions, and diseases. These studies required extraction of appropri-
ate data sets from multiple seemingly unrelated sources, their compilation or
customization, and the application of data analysis and/or data visualization
tool resources to these data sets. Such operations are inherently performed in an
ad hoc manner, and should be executable rapidly enough to follow up on fleeting
ideas.
The creation of information systems that may work as a science infrastruc-
ture to accelerate further development of science and technology requires a new
generic technology for ‘knowledge federation’, as well as the development and
extension of databases, simulators, and analysis tools in each research field. For
the routine jobs that require interoperation of multiple knowledge resources, we
may use work-flow and/or resource-orchestration technologies based on GRID
computing. These technologies were extensively studied and developed during
the last decade in the US, Europe and Japan, and are now becoming off-the-shelf
technologies. But in R&D activities, where repetitive trial-and-error operations
are a vital part of creative thinking, we need support for ad hoc, instantaneous
implementation of passing ideas. By contrast, most of the preceding e-science
projects were based on GRID computing technologies that are suited only to
routine work.
Ad hoc knowledge federation is required in strategic intellectual activities
including not only scientific research activities, but also strategic analysis and
planning such as financial analysis and planning, and disaster contingency anal-
ysis and planning.
Recently, the Web works not only a world-wide repository of documents, but
also an infrastructure to make mobile phones, sensor/actuator networks, home
electronic appliances, and expensive scientific equipments accessible to and from
Web resources. Users can access and/or control these objects through the Web
from anywhere, and these objects can get useful information from anywhere
through the Web, and/or update some Web resources. Figure 1 shows the current
situation of IT environments. The Web provides a basic network of knowledge
resources. Mobile phones are connected to some Web servers. A mobile phone
can access any Web resources via such a server. It can be also accessed from
some Web resources via such a server. A sensor/actuator network with ZigBee
protocol can dynamically span a network of a coordinator and sensor/actuator
nodes. The coordinator can read the value of any sensor, and/or write the value
of any actuator, in any node. By defining a Web application or a Web service for
the computer with this coordinator node, we can make the value of each sensor
of every node accessible through this Web application or Web service. In each
house, home electronic appliances are connected to a home network, which is
connected to the Internet through a network node. Therefore, these appliances
are accessible to and from the Web for their monitoring and control. Figure1 also
shows ad hoc peer-to-peer networks without any connection to the Web. Nodes
in these networks cannot access nor be accessed by any Web resources.
All different types of functions that are provided by mobile devices, sen-
sor/actuator nodes, and electronic appliances are accessible through the Web as
Web resources, namely as Web applications or Web services. Therefore, we may
consider them as knowledge resources. We can apply our knowledge federation
technology to extract some of their sub functions, to make them interoperate
Web
Web-based access
Mobiles
Coordinator
e.g.ZigBee
Network proximity-based
Sensor networks access
Fig. 1. Current IT environments consisting of the Web, mobile networks, and sensor
networks.
with each other, and to compose a new knowledge resource as a Web application
or a Web service.
4.2 Meme media as enabling technologies
We have been conducting the research and development of ‘meme media’ and
‘meme market’ architectures since 1987. The word ‘meme’ was coined by Richard
Dawkins in 1976 [28]. He pointed out the similarity between biological evolution
and cultural evolution. Ideas are replicated and propagated from people to peo-
ple, which corresponds to genetic replication. Two different ideas are recombined
to produce a new idea, which corresponds to genetic recombination. An idea may
be propagated from a person to another with some error, and the resultant idea
may become a new useful idea, which corresponds to genetic mutation. Finally,
ideas are evaluated by a community, and only those that are frequently referred
to remain as useful ideas. Dawkins coined the word ‘memes’ by combining two
words; gene and mimesis. A meme denotes a virtual entity that carries a unit of
knowledge, as a gene works as a unit of carrying genetic information.
A meme media architecture denotes a component-based knowledge media ar-
chitecture in which users can visually combine meme media objects to compose
a new meme media object, and decompose a composite one to reuse its compo-
nent meme media objects for different compositions. It enables users to replicate
and reedit knowledge resources on meme media, and to distribute them among
people. If a world-wide repository of meme media objects is provided, knowledge
resources that are published into this repository as meme media objects are repli-
cated and recombined by a huge number of people through direct manipulations,
and naturally selected by user communities. Such a repository works as a meme
pool. A meme pool promotes the evolution of knowledge resources published as
meme media objects.
The current Web works as a publishing repository of knowledge resources,
but does not work as a meme pool since users cannot easily extract fragments
of Web resources and recombine them to compose a new Web resource only
through direct manipulation of Web resources without writing any codes. Some
Web mashup tools could be extended to share the same goal with meme media
architectures, but they are still premature for the above mentioned purpose.
We developed the first versions of 2D and 3D meme media architectures
‘IntelligentPad’ [29] and ‘IntelligentBox’ [30] respectively in 1989 and in 1995,
and have been working on their meme-pool and meme-market architectures,
as well as on their applications and revisions. IntelligentPad represents each
component as a pad, a sheet of paper on the screen. A pad can be pasted on
another pad to define both a physical containment relationship and a functional
linkage between them (Figure 2). When a pad P2 is pasted on another pad P1,
the pad P2 becomes a child of P1, and P1 becomes the parent of P2. No pad
may have more than one parent pad. Pads can be pasted together to define
various multimedia documents and application tools. Unless otherwise specified,
composite pads are always decomposable and reeditable.
P2 P3
P1
P3 P1
P2
Fig. 2. A composite pad and slot connections.
In object-oriented component architectures, all types of knowledge fragments
are defined as objects. IntelligentPad exploits both an object-oriented compo-
nent architecture and a wrapper architecture. Instead of directly dealing with
component objects, IntelligentPad wraps each object with a standard pad wrap-
per and treats it as a pad (Figure 2). Each pad has both a standard user interface
and a standard connection interface. The user interface of a pad has a card like
view on the screen and a standard set of operations like ‘move’, ‘resize’, ‘copy’,
‘paste’, and ‘peel’. Users can easily replicate any pad, paste a pad onto another,
and peel a pad off a composite pad. Pads are decomposable persistent objects.
You can easily decompose any composite pad by simply peeling off the primi-
tive or composite pad from its parent pad. As its connection interface, each pad
provides a list of slots that work in a similar way as connection jacks of an AV-
system component, and a single connection to a slot of its parent pad (Figure 2).
Each pad uses a standard set of messages ‘set’ and ‘gimme’ to access a single slot
of its parent pad, and another standard message ‘update’ to propagate its state
change to its child pads. In their default definitions, a ‘set’ message sends its
parameter value to its recipient slot, while a ‘gimme’ message requests a value
from its recipient slot. Figure 3 shows a set of pulleys and springs that are all
represented as pads, and a composition with them. Each of these pulleys and
springs is animated by a transparent pad. The old version of IntelligentPad could
not directly deal with a graphical object animated on a canvas as a pad. Each
pad was required to have a rectangular canvas. This restriction has been recently
removed in the 2008 version of IntelligentPad that has employed SVG and runs
on advanced de facto standard browsers empowered by Spaceflight plug-in.
(a) primitive pads (b) a composite pad
Fig. 3. An example pad composition.
4.3 Wrapping Web resources with 2D meme media for knowledge
federation
From 2003 to 2004, we developed new technologies for users to easily wrap Web
applications and Web services into meme media objects [14, 15, 31]. C3W [15]
is a wrapper framework that enables users to open a browser showing a Web
application page, to clip out some input forms and some output portions as
pads, and to make such pads, that are clipped out from more than one Web
page, interoperate with each other. C3W enables users to make these definitions
only through direct manipulation without writing any program codes. For a
given Web service with a WSDL description of its interface, our Web service
wrapper enables users to define its proxy object as a pad with a specified subset
of IO signals. Our Web service wrapper first pops up the IO signal list of the
Web service for users to arbitrarily select some of them as the slots of the pad
to define, and to specify default values for some other input signals. Users can
make these specifications only through direct manipulation. Then the wrapper
automatically creates a pad that wraps the original Web service. This pad works
as a proxy object of this Web service.
Figure 4 shows, at its bottom right corner, a Web application by US Naval
Observatory showing day and night over the earth. The left object is a composite
pad showing the difference of arbitrarily chosen two seasons; the same time on
summer solstice and winter solstice for example. This was constructed by just
clipping out the date input form and the simulated result as pads from the
Naval Observatory Web application, drag-and-dropping them on a special pad,
and applying the multiplexing to the input to obtain multiple outputs.
Fig. 4. Construction of a composite pad, using clips extracted from Web pages as pads.
All these operations were performed through direct manipulation of pads. For
browsing the original Web page, we use a Web browser pad, which dynamically
frames different extractable document portions for different mouse locations so
that its user may move the mouse cursor around to see every extractable doc-
ument portion. When it frames a desired object, you can just drag the mouse
to clip out this object as a pad. All the pads thus clipped out from Web pages
in a single navigation process keep their original IO functional relationship even
after their arrangement on the same special pad C3WSheet. The C3WSheet pad
gives a cell name to each pad pasted on it. The pads clipped out from date input
forms are named as E, F, G, H, I cells. They respectively correspond to year,
month, day, hour, and minute input forms. The simulation output pad is named
as J cell. Whenever you input a new date to the extracted date input forms, the
corresponding extracted output pad showing a simulated result will change its
display. This simulation is performed by the server corresponding to the source
Web application from which all these pads are extracted.
The multiplexer pad, when inserted between the base C3WSheet pad and
the extracted date input form pad, automatically inserts another multiplexer
pad between the base C3WSheet pad and every extracted pad whose value may
depend on the input to this date input form. If you make a copy of the extracted
date input form pad on the multiplexer pad, each copy of this multiplexer pad
automatically makes a copy of its child pad that was also extracted from the
same Web application and is dependent on the update of the date value. Mutually
related multiplexer pads maintain the relationship among the copies of their child
pads. The original copies, the second copies, and the third copies respectively
form independent tuples of extracted pads, and each tuple maintains input and
output relationship among its constituent pads. In Figure 4, two simulation
results are obtained for two different days.
The above mentioned composite pad allows us to specify any time in GST.
You can easily modify this content to accept any local time by combining this
with another public Web application for time conversion. The C3WSheet pad in
Figure 4 has four more multiplexed pads named as A, B. C. and D cells. They
are extracted from the time conversion service. The pads named as A, B, C cells
are the extracted input forms, respectively specifying the input local time, the
source time zone, and the destination time zone. The pad named as D shows
the local time in the destination time zone. The destination time zone is fixed to
GST in the cell C. You can just specify equations in the cells E, F, G, H, and I
to make their values always equal to the year, month, day, hour, and minute of
the cell value of D. Such a direct manipulation process customizes existing Web
resources, and makes some of them interoperate with each other to define a new
knowledge resource as a composite pad.
Figure 5 shows our basic idea of wrapping different types of knowledge re-
sources into meme media objects for achieving interoperability among them. The
arrows (1) and (2) respectively show the wrapping of Web applications and Web
services into meme media objects. The arrow (6) denotes the combination of
more than one meme media object by direct manipulation to compose a new
composite meme media object. The arrow (5) denotes the wrapping of legacy
applications into meme media objects. Legacy applications without GUI can
be always easily wrapped into meme media objects if they expose their APIs.
Legacy applications with GUI are not always easily wrapped into meme media
objects even if they expose their APIs. The arrow (3) denotes that any composite
pad can be accessed as a Web service through SOAP protocol. We have already
developed a tool that automatically creates a corresponding Web service for any
given composite meme media object. The arrow (4) denotes the conversion of an
arbitrary composite meme media object into a Web application. This conversion
is called ‘flattening’ since it converts a composite media object into a flat Web
page viewable with any Web browser. Since composite meme media objects are
not HTML documents, they are not directly viewable with Web browsers. We
have already developed a tool for the automatic flattening of composite meme
media objects. This mechanism however makes it easy to develop phishing Web
sites just by pasting a password stealing pad over a wrapped trustworthy Web
application to define a composite pad. Because of the inherent fragility of the
current Web technology against phishing trials, we have made a decision that
we should not provide this flattening conversion mechanism to the public.
The Web
Web
applications Web services
(1) (2)
(4) (3) Legacy
wrapping Applications
Meme Media Objects (5)
Composition
(6)
Meme Media and a Meme Pool t
Fig. 5. A basic architecture to wrap different types of knowledge resources into meme
media objects for their mutual interoperability.
While the flattening mechanism enables us to use the Web itself as a meme
pool, we developed another world-wide repository Piazza that works as a meme
pool for pads. Piazza was developed on top of Wiki, and uses Wiki servers.
Piazza provides its own browser to access different pages. Each page that can
be identified by a URL may contain arbitrary number of pads. Users can open
an arbitrary Piazza page, and drag and drop pads between this page and a local
pad environment. By clicking the page registration button, users can reflect the
page update to the servers. By clicking the page refresh button, users can reflect
the current state of the server to the displayed Piazza page. We have already
shown that Piazza can be easily developed by applying C3W framework to Wiki
system [15]. Figure 6 shows the Piazza browser displaying a Piazza page with
several registered pads.
In order to make a meme pool Piazza system from Wiki, you can first access a
Wiki page, and clip out the URI input, the HTML input form, the refresh button,
and the output page as pads, and paste them on the same C3WSheet pad. You
need to paste a PadSaverLoaderPad as a cell of the same C3WSheet pad, and
relate its input and output respectively to the extracted input form pad and
Fig. 6. A worldwide repository of pads developed by applying C3W framework to Wiki.
the extracted output page. A PadSaverLoaderPad makes conversion between
a pad on itself and its save format representation in XML. Suppose that the
PadSaverLoaderPad, the extracted HTML input form pad, and the extracted
output page pad are assigned to cells A, B, and C. The relationship among
them is defined as follows. We define the equation for the cell A as ←C, and the
equation for the cell B as ←A. People can access any page specifying its URI,
drag-and-drop arbitrary composite pads to and from the PadSaverLoaderPad of
the composed pad to upload and download them to and from the corresponding
Wiki server. Each page is shown by the PadSaverLoaderPad. For a jump from a
page to another page in a Piazza system, we can use an anchor pad that can be
pasted on a Piazza page. This anchor pad holds a URI that can be set through
its #refURI slot, and, when it is clicked, sets this URI to the #URI slot of the
Piazza, i.e., to the #URI slot of its base C3WSheet pad.
Piazza allows people not only to publish and share knowledge resources rep-
resented as pads, but also to compose new knowledge resources by combining
components of those pads already published in it. People can publish such newly
composed knowledge resources into the same Piazza system. The collaborative
reediting and redistribution of knowledge resources in a shared publishing repos-
itory by a community or a society of people will accelerate the memetic evolution
of knowledge resources in this repository, and make it work as a meme pool.
4.4 Knowledge federation and the memetic Web
The most recent 2008 version of IntelligentPad fully exploits Microsoft Silverlight
technology, and runs on Internet Explorer 6 and 7, Mozilla Firefox, or Safari
browser empowered by Silverlight plug-in. It is a Web-top system that allows us
to directly manipulate pads on a Web page.
It is one important feature of this version that its pad may not have its
rectangular canvas. Any graphical object including a line segment can be treated
as a pad (Figure 7). Three needles of the analog clock in this figure are all pads.
Such a needle does not need a transparent canvas on which it is animated. Such
a pad is called a canvas-free pad. Our preceding versions could not deal with
canvas-free pads. This new version exploits SVG (Scalable Vector Graphics) to
describe graphical objects. When a pad is pasted on a canvas-free pad to work as
its child pad, the former is bound by the local coordinate system spanned by the
latter pad, instead of being clipped by the canvas area of the parent pad. This
version also provides various primitive pads for the linear transformation of local
coordinate systems. Figure 7 (b) shows an application of an affine transformation
to the same gear complex in Figure 7 (a).
(a ) a composition using a n a na log clock a nd a gea r complex, with the dia l pla te being ma de
tra nslucent.
(b) a n a ffine tra nsforma tion a pplied to a gea r complex
Fig. 7. The canvas-free IntelligentPad, and some of its example pads.
It is another important feature of this new version that we need no Intel-
ligentPad kernel running on clients. Any advanced de facto standard browser
empowered by Silverlight plug-in can provide a Web-top environment of pads,
and render any composite pad as an extended Web page. Users accessing through
such browsers can directly manipulate pads on a Web page, and do not recog-
nize any difference between Web resources and composite pads. Composite pads
can be also registered in HTTP servers for other people to access them as Web
documents.
The 2008 version of IntelligentPad, together with an advanced de facto stan-
dard Web browser empowered by Silverlight plug-in, extends the Web to ‘the
memetic Web’ as shown in Figure 8. The memetic Web allows people to publish
not only Web resources, but also composite pads. People can browse both of
them just by using a browser. Client PCs need not be installed with an Intelli-
gentPad kernel system. Users can extract resource fragments as pads from Web
resources through direct operation by using C3W technology. They can also ac-
cess a composite pad, and make copies of any of its component pads. They can
combine these pads extracted from the memetic Web, as well as locally available
pads, to compose a new composite pad for their reuse, and publish it into the
memetic Web for the further reuse by other users. The memetic Web works as
a meme pool of both Web resources and composite pads.
Legacy
Applications
Web proxy pads
Services Web
(2) service A
(5) reediting &
Web
(3) The Web
redistribution
service B
Web (1) Composite
Applications Pads and
Boxes clipping
Compositon and (6) RIA (Silverlight) tech. copy
Orchestration
The memetic Web reuse in local
publish
The Memetic Web
environments Only requires Silverlight-empowered browser!
Fig. 8. The memetic Web and different types of knowledge resources.
The handling of more than one pad extracted from different Web applications
on the same web top requires so-called cross-frame scripting, which is not allowed
by any browser. To solve this problem, the memetic Web uses a proxy server
called a C3W server through which each pad of this type accesses the original
Web application. The access of different Web applications through the same
proxy server solves the cross-frame scripting problem.
The memetic Web allows people to define both service federation and view
federation of knowledge resources instantaneously in an ad hoc way just through
direct manipulation of Web resources and pads. While mashup technologies focus
on composing a new Web resource from available Web resources, the memetic
Web focuses on the repetitive reediting, reuse, and republication of knowledge
resources by end users in an instantaneous and ad hoc way.
5 Concluding Remarks
This paper focused on knowledge federation from different aspects; its taxon-
omy, its necessity, and its required technologies. First, it clarified three different
aspects of knowledge federation, namely, semantic federation, service federation,
and view federation. Then it focused on ad hoc knowledge federation of Web re-
sources including Web applications. Ad hoc knowledge federation deals with two
of the three aspects, i.e., service federation and view federation. Then it reviewed
the current author’s R&D on this type of knowledge federation, and proposed
an extension of the Web toward the memetic Web. It also introduced our new
version of the IntelligentPad system as an enabling technology for this exten-
sion. This version runs on de facto standard browsers empowered by Silverlight
plug-in, and requires no IntelligentPad kernel running on clients.
The memetic Web will become a next-generation Web medium for re-editable
knowledge resources, and for user-centric ad hoc knowledge federation of knowl-
edge resources. It is accessible to anyone with a mainstream Web browser. The
memetic Web will extend the power of community editing to the deep Web
storing not only static resources but also executable resources.
The current author is especially interested in its application to accelerate
the evolution of science and technology. Current e-science and cyber infrastruc-
ture projects are all based on GRID computing. The GRID computing pro-
vides HPC (High-Performance Computing) power, services, and workflow def-
inition/execution/management systems for service composition. It emphasizes
definition, execution, and management of routine or frequently requested jobs,
but not user-centric, immediate, unforeseen ad hoc composition with existing
resources. It provides tools for editing internal representations of service compo-
sitions, but no tools for directly reediting compound documents to compose new
knowledge resources. GRID-based service composition provides unified compo-
nent architecture for both GRID services and Web services, but cannot deal with
Web applications. When a new service composition becomes available to public
users, the task is already well known to almost every researcher as a routine or
frequently used method. How can users define and execute an untried new com-
bination of knowledge resources, for example, to discover interesting correlations
between the El Niño-Southern Oscillation and the weather in Europe, volcanic
eruptions, and diseases? We want to enable such trials to be set up within 10-
20 minutes so that fleeting new ideas can be immediately tried and evaluated.
GRID-based e-science and cyber infrastructure projects do not answer this ques-
tion. Ad hoc knowledge federation will become a potential enabling technology
to satisfy such requirements in science and technology.
References
1. Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American
(May 2001) 34–43
2. Biezunski, M., Newcomb, S., Pepper, S.: ISO/IEC 13250:2002, Topic Maps. PDF
format (2002)
3. Dean, M., Connolly, D., van Harmelen, F., Hendler, J., Horrocks, I., McGuinness,
D., Patel-Schneider, P., Stein, L.: OWL Web Ontology Language 1.0 Reference.
Cambridge, MA: World Wide Web Consortium (W3C) (www.w3.org/TR/owl-
features) (2003)
4. Dou, D., Mcdermott, D.V.: Ontology translation by ontology merging and auto-
mated reasoning. Yale University, New Haven, CT (2004)
5. Horrocks, I.: DAML+OIL: A Reason-able Web Ontology Language. In: Pro-
ceedings of the 8th International Conference on Extending Database Technology:
Advances in Database Technology. (March 2002) 2–13
6. Baird, D.: Thing Knowledge: A philosophy of scientific instruments. University of
California Press (2004)
7. Stefik, M.: The next knowledge medium. The AI Magazine 7(1) (1986) 34–46
8. Heimbigner, D., McLeod, D.: A federated architecture for information manage-
ment. ACM Trans. Inf. Syst. 3(3) (1985) 253–278
9. Edwards, W.K., Joy, B., Murphy, B.: Core JINI, Prentice Hall Professional Tech-
nical Reference. (2000)
10. Gelernter, D.: Generative communication in linda. ACM Trans. Program. Lang.
Syst. 7(1) (1985) 80–112
11. Picco, G., Murphy, A., Roman, G.: Lime: Linda meets mobility. In: ICSE 99: Pro-
ceedings of the 21st international conference on Software engineering, Los Alamitos,
CA, USA, IEEE Computer Society Press (1999) 368–377
12. Sun Microsystems: Javaspaces service specification, version 1.2 (2002)
13. Sun Microsystems: Jini technology core platform specification, version 1.2 (2001)
14. Ito, K., Tanaka, Y.: A visual environment for dynamic web application compo-
sition. In: HYPERTEXT ’03: Proceedings of the fourteenth ACM conference on
Hypertext and hypermedia, New York, NY, USA, ACM (2003) 184–193
15. Tanaka, Y., Ito, K., Fujima, J.: Meme media for clipping and combining web
resources. World Wide Web 9(2) (2004) 175–184
16. Tanaka, Y.: Knowledge federation over the web based on meme media technologies.
In: Lecture Notes in Computer Science, 3847. (2006) 159–182
17. Agrawal, R., Bayardo, R., Jr., D.G., Papadimitriou, S.: Vinci: a service-oriented
architecture for rapid development of web applications. In: WWW ’01: Proceedings
of the 10th international conference on World Wide Web. (2001) 355–365
18. DBpedia. http://dbpedia.org/About
19. Maamar, Z., Benslimane, D., Ghedira, C., Mahmoud, Q.H., Yahyaoui, H.: Tuple
spaces for self-coordination of web services. In: SAC ’05: Proceedings of the 2005
ACM symposium on Applied computing. (2005) 1656–1660
20. Sotomayor, B., Childers, L.: Globus Toolkit 4: Programming Java Services. Morgan
Kaufman, San Francisco, USA (2006)
21. Emmerich, W., Butchart, B., Chen, L., Wassermann, B., Price, S.: Grid service
orchestration using the business process execution language (BPEL). Grid Com-
puting 3(3–4) (2005) 283–304
22. Mcllraith, S., Son, T.C., Zeng, H.: Semantic web services. Intelligent Systems
16(2) (2001) 46–53
23. Tanaka, Y.: Meme Media and Meme Market Architectures: Knowledge Media for
Editing, Distributing, and Managing Intellectual Resources. IEEE Press & Wiley-
Interscience, NJ (2003)
24. Young, M.: Google Maps Mashups with Google Mapplets (Firstpress), 1 edition.
APress (2008)
25. R..Ennals, Gay, D.: User-friendly functional programming for web mashups. SIG-
PLAN Not. 42(9) (2007) 223–234
26. Ankolekar, A., Krötzsch, M., Tran, T., Vrandecic, D.: The two cultures: mashing up
web 2.0 and the semantic web. In: WWW ’07: Proceedings of the 16th international
conference on World Wide Web. (2007) 825–834 SESSION: Semantic web and web
2.0.
27. Tanaka, Y., Fujima, J., Ohigashi, M.: Meme media for the knowledge federation
over the web and pervasive computing environments. Advances in Computer Sci-
ence – ASIAN 2004, Lecture Notes in Computer Science 3321 (2004) 33–47
28. Dawkins, R.: The Selfish Gene. Oxford Univ. Press (1976)
29. Tanaka, Y., Imataki, T.: IntelligentPad: A hypermedia system allowing functional
compositions of active media objects through direct manipulations. In: Proceedings
of the IFIP 11th World Computer Congress. (1989) 541–546
30. Okada, Y., Tanaka, Y.: Intelligentbox: a constructive visual software development
system for interactive 3d graphic applications. In: Proc. of the Computer Anima-
tion 1995 Conference, Los Alamitos, CA, USA, IEEE Computer Society (1995)
114–125
31. Tanaka, Y.: Knowledge media and meme media architectures from the viewpoint
of the phenotype-genotype mapping. In: SIGDOC ’06: Proceedings of the 24th
annual ACM international conference on Design of communication, New York,
NY, USA, ACM (2006) 3–10