Stephane Boyera Philippe Cudr´e-Mauroux Mathieu D’Aquin J´erˆome David Victor De Boer Christophe Gu´eret Florent Pigout Mike Powell Louiqa Raschid Laurens Rietveld Stefan Schlobach Elena Simperl Andreas Thor 1 6 14 23 31 39 40
Guillermo Gonz´alez-Moriy´on1, Emilio Rubiera1, Marcos Sacrist´an2, and Javier
Collado2 1 Fundacio´n CTIC
Gijo´n, Asturias, Spain {name.surname}@fundacionctic.org http://www.fundacionctic.org/
2 Treelogic
Llanera, Asturias, Spain {name.surname}@treelogic.com http://www.treelogic.com/ 1
Near 100,000 people died in Europe in the first decade of this century due to
natural and industrial disasters [
In the current era, developed countries have a number of bodies that respond to emergencies in order to minimize casualities and economic loss. First responders include fire brigades, police, hospitals and military forces. An effective response depends on their ability to quickly take decisions based on accurate and reliable data. Emergency Management Systems (EMS) are information tools supporting resolutive actions against the clock.
Nowadays Europe showcases a well-spread heterogeneity of EMSs depending on the stakeholder “color” (emergency jargon to identify first responders) and the administrative division (city, region, etc.) where they operate. Political boundaries and compartimentation of functional competences are a hindrance for fluid communication and information exchange between EMSs. This fragmentation has evolved with time for cultural and historic reasons, and it is an unavoidable part of the European reality. The complexity of some emergency situations requires the participation of a variety of first responders who depend crucially on their ability to effectively exchange data (Figure 1). 2
At the event of a disaster many entities are involved. Some may be affected by the disaster, others may need to take contigence actions, while others may be accidental watchers of the incident. Thus, each entity can provide a partial version of the total picture. The objective is to make the most complete picture available to decision makers by putting together the partial pictures coming from the different parts involved. Due to the decentralization of the information scenario a number of conditions must be fulfilled: – It must be possible to gather and to manage heterogeneous information from many available sources. This information includes an appropriate description of the disaster event in terms of both quality and quantity. In addition, the description of the scenario must be enriched with contextual and environmental data, e.g., infrastructures, populated areas, weather forecast... This harvest process often faces challenges due to the incompleteness and incorrectness of the data. – The description of the scenario must be shared and agreed by the stakeholders in real time. A common picture facilitates team coordination and synchronization of response operations, for instance, to know in real time which areas have been already evacuated. – Information overload must be prevented. Decisions are difficult to make when dealing with an overwhelming volume of heterogeneous data. Filtering information in terms of relevance is crucial. Moreover, the concept of relevance is subjective: each actor is interested in a different subset of the information. – Information pieces must be referenced with respect to both geographical and temporal coordinates. To gain an insight of the situation, it helps to have map interfaces and events displayed in sequence. – Message oriented communications among different responders must be effective regardless of cultural and technical differences. Due to the diversity at each end of the channel, many problems arise including divergences of communication protocols, data formats or information models. In addition to the core message, the sender may decide to include additional information that enriches the message and increases its usefulness to the receiver. – When integrating information from various sources and communication breakdown occurs, previously exchanged information might still be valuable. Not all data has the same expiration date: topographic information will still be valid, whereas current positioning of units deployed is dynamic and potentially untrustable afterwards.
The aforementioned conditions match some of the research topics tackled by the Semantic Web, e.g., information quality, data and model sharing, or data enrichment. These matchings suggest that a potential solution to the EMS interoperability problem may reuse the findings from the Semantic Web community. Moreover, the work on Semantic Web has demonstrated the value of adopting open standards to reuse vocabularies and exchange data between decoupled systems in the absence of a central authority. 3
DISASTER (Data Interoperability Solutions at STakeholders Emergencies Reaction) is a collaborative European project including software vendors, research organizations and specialists in emergency response. The goal of DISASTER is to ease communication among existing EMSs considering the requirements aforementioned.
DISASTER project respects EMSs diversity in Europe. No “one size fits all” central EMS or unique exchange language is proposed. Instead, DISASTER relies on a shared reference ontology as well as on mediation techniques. As depicted on Figure 2, semantic technologies will make possible for any EMS to exchange information and to query external sources including Geographical Information Systems and the Linked Open Data cloud, and of course, other EMSs.
When it comes to mutual understanding, the first step is to agree on shared
baseline knowledge. Under our approach, this knowledge is modelled with
ontologies. Ontologies have been proved to solve common understanding problems [
Moreover, the use of ontologies to specify knowledge in the domain of
emergencies have already been discussed in a number of studies [
DISASTER has commenced in February 2012 and will run until 2015. The project has started by gathering requirements about different aspects: from first responders operative requirements to linguistic and cultural requirements.
DISASTER proposes data mediation to tackle communication problems among EMSs. Efforts are planned to bring to practice the theoretical results of the research community on data mediation. Actual interoperability among EMSs poses a number of challenges regarding several dimensions of the problem: communication protocols, data models and data formats. A workable solution typically involves all these aspects.
The main result of the project will be the DISASTER ontology. The
creation of this ontology will follow a modular approach: the core module will be
based on upper level ontologies such as DOLCE [
The ontology will serve to combine heterogeneous information coming from diverse sources. The project will study how to coherently assemble an agreed disaster scenario description. Moreover, map-based visualization paradigms will be explored to merge context information with operational data and user-oriented mechanisms will be defined to filter relevant subsets of information. Devices supporting on site operations should be taken into account. Responders are often equipped with mobile devices providing access to the information. However, connectivity issues must be addressed as connection cannot be ensured.
Finally, in order to ensure the DISASTER solution is grounded on reality, validation will be carried out. At this early stage of the project the details of the final project scenario are still a draft. However, one of the possibilities involves a (simulated) disaster at a large international European airport with a transnational component featuring stakeholders of different kinds and nationalities.
The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n 285069.
Linked data from your pocket Jérôme David, Jérôme Euzenat, Maria-Elena Ro¸soiu
INRIA & Pierre-Mendès-France University
Grenoble, France {Jerome.David,Jerome.Euzenat,Maria.Rosoiu}@inria.fr Abstract. The paper describes a lightweight general purpose RDF framework for Android. It allows to deal uniformly with RDF, whether it comes from the web or from applications inside the device. It extends the Android content provider framework and introduces a transparent URI dereferencing scheme allowing for exposing device content as linked data. 1 Introduction Smartphones are becoming the main personal information repositories. Unfortunately, this information is stored in independent silos managed by applications and thus, it is difficult to share data across them. Nowadays, mobile operating systems, such as Android, deliver solutions in order to overcome this, but they are limited by the application database schemas that must be known beforehand.
The difficulty to share phone data at the web scale can be seen as another drawback. One can synchronize application data, such as the contacts or the agenda using a Google account. However, they are not generic solutions and there it is no mean to give access to data straight from the phone.
Our goal is to provide applications with a generic layer for data delivery in RDF. Using this solution, applications can exploit device information in an uniform way without knowing from the beginning application schemas. This information can also be exposed to the web and web information can be considered in the same uniform manner. Moreover, we propose to do it along the linked data principles (provide RDF, describe in ontologies, use URIs, links to other sources).
For example, in the future, this application could be used as a personal assistant. When one would like to know which of his contacts will participate to a scientific conference, he can query the calendar of all his contacts in order to retrieve the answer. For sure, according to the security settings of the corresponding contact, he may be allowed or not to access the calendar.
The mobile device information can as well be accessed remotely, from any web browser, by any persons who has granted the access to it. In this case, acts like a web server.
We presented a first version of the RDF content provider in [
We first describe the context in which the Android Platform stores its data, and how it can be extended in order to integrate RDF. Then, we present two applications that sustain its feasibility: the first one is an RDF browser that acts like a linked data client and allows the navigation through the device information, and the second one is an RDF server which exposes its information to the outside world. We continue to present the challenges raised by such applications and solutions we implemented for them. Finally, we conclude presenting future improvements and challenges in this field. 2
Android Content Providers Inside the Android system, each application runs in isolation from other applications. This Linux-based operating system assigns to each application a different and unique user. Only this user is granted access to the application files. This allows one to take advantage of a secure environment, but prevents the exchange of data across applications. To overcome this drawback, Android provides the content provider mechanism.
Content providers enable the transfer of structured data between device
applications. They encapsulate the data and control the access to it through an interface. This
interface empowers one to query the data or to modify it ([
A content provider is a subclass of ContentProvider and implements the following interface: Cursor query( Uri id, String[] proj, String select, String[] selectArgs, String orderBy ) Uri insert( Uri id, ContentValues colValueList) int update( Uri id, ContentValues colValueList, String select, String[] selectArgs ) int delete( Uri id, String select, String[] selectArgs ) String getType( Uri id ) .
With the content provider API, each data (table or individual) is identified by a URI having the following structure: The content: scheme is the cornerstone of each Content Provider URI, the authority identifies the provider, i.e., the dataset, and the path/to/data identifies a particular table or individual (row) in the dataset. For example, the URI content://contacts/people refers to all the people in the contact application, and the URI content://contacts/people/33 identifies a specific instance of these, namely the instance having the id 33.
When an application wants to access a particular piece of data, it queries its URI. This is done through a call to the ContentResolver which is able to route the query to the right content provider.
From a semantic web point of view, using URIs to identify data is a strong point of the Android content providers. Still, there are several limitations if we would like to use them as a linked data interface.
Specifically, URIs used by content providers are local to each device, i.e., not dereferenceable on the web, and not unique. The content scheme used by providers is not a standard protocol. Furthermore, two distinct devices will use the same URI to identify different data. For example, by using content://contacts/people one would be able to access the contacts from both devices.
Another drawback is the SQL interface of the Android content providers. The queries are issued in an SQL manner and the results are presented to the user as a table. 3
The RDF Content Provider Framework We designed the RDFContentProvider framework to give a semantic web flavour to Android and to overcome these problems. It is composed of the RDFProvider API and the RDFContentResolver application. The API must be included inside the applications that want to access RDF providers and inside the applications that want to define new RDF content providers. TheRDFContentResolver application is the one that records all the RDF content providers installed on the device and routes queries to the relevant provider. Figure 1 gives an overview of the framework architecture.
RDFHttpContentProvider RDFPhoneSensorsContentProvider
RDFBrowser uri uri uri rdf rdf
uri RDFContentResolver
u rdf ri r d f uri – RdfContentProviderWrapper: A subclass of RdfContentProvider which allows for adding RDF content provider capabilities to an existing classical content provider.
RDFContentProvider follows primarily the same kind of interface as ContentProvider. The minimal interface to implement linked data applications is: – RDFCursor getRdf( Uri id ) The Cursor iterates on a table of subject-object-predicate (or object-predicate) which are the triples involving the object given as a URI. If one wants to offer a more elaborate semantic web interface, i.e., a minimal SPARQL endpoint, the following methods have to be also implemented: – Uri[] getTypes( Uri id ): returns the RDF types of a local URI; – Uri[] getOntologies(): ontologies used by the provider; – Uri[] getQueryEntities(): classes and relation that the provider can deliver; – Cursor query( SparqlQuery query ): returns results tuple; – Cursor getQueries(): triple patterns that the provider can answer.
The RDF providers that we have developed so far are implementing only the first three primitives. 3.2
The RDF Content Resolver Service The RDFContentResolver service has the same goal as the ContentResolver belonging to the Android framework. It maintains the list of all the installed RDF content providers, and forwards the queries it receives to the corresponding one. This application is never visible to the user, therefore we have implemented it as an Android service.
When an RDF Content Provider is instantiated by the system, this provider automatically registers to the RDFContentResolver. A principle similar to the one from the Android Content Provider framework is used.
The RDFContentResolver can route both the local (content:) and external (http:) URI-based queries. In case of a local URI, i.e., starting with the content scheme, the resolver decides to which provider it must redirect the query. In case of an external URI, i.e., starting with the http scheme, the provider automatically routes the query to the RDFHttpContentProvider (see Figure 1).
The RDFHttpContentProvider allows one to retrieve RDF data from the Web. It parses RDF documents and presents them as RDFCursors. So far, only the minimal interface has been implemented, i.e., the getRdf( Uri id ) method. 3.3
RDF Providers for Address Book, Calendar and the Phone Sensors The RDF Content Resolver application is also bundled with several RDF content providers encapsulating the access to Android predefined providers. The Android framework has applications that can manage the address book and the agenda. These two applications store their data inside their own content provider.
In order to expose this data as RDF, we developed the RDFContactProvider and the RDFCalendarProvider. These providers are wrapper classes for the ContactProvider and the CalendarProvider residing inside the Android framework.
RDFContactProvider exposes contact data using the FOAF ontology. It provides data about a person’s name (display name, given name, family name), about its phone number, email address, instant messenger identifiers, homepage and notes.
RDFCalendarProvider provides access to the Android calendar using the RDF Calendar ontology1. The data supplied by this provider contains information about events, their location, their date (starting date, ending date, duration, and event time zone), the organizer of the event and a short description.
RDFPhoneSensorsContentProvider aims to expose sensor data from the sensors embedded inside the mobile device. Contrary to the others, they are not offered as Content Providers. At the present time, it only delivers the geographical position (retrieved using the Android LocationManager service). In order to express this information in RDF, we use the geo location vocabulary2, the one that provides a namespace for representing lat(itude) and long(itude). 4
RDF Browser The RDF Browser acts like a linked data client. Given a URI, the browser makes an HTTP URI request in order to retrieve the information from the specified location. If the data contains other URIs, the user can click on them and the browser will issue a new query with this URI.
An example can be found in Figure 2. In this case, the user uses the RDFBrowser to get the information about the contact having the id 4. When the browser receives the request, it sends it further to the RDFContentResolver. Since the URI starts with the content:// scheme and has the com.android.contacts authority, the resolver routes the query to the RDFContactProvider. This provider retrieves the set of triples describing the contact and sends it to the calling application which displays it to the user. Thereupon, the user decides that he wants to continue browsing and selects the contact’s homepage. In this case, since the URI starts with the http:// scheme, the resolver routes the query to the RDFHttpContentProvider. The same process repeats and the user can see the remote requested file, i.e., Tim Berners-Lee FOAF file. 5
RDF Server The RDF Server is a new component added to the architecture. This server provides to the outside world the data stored into the device as RDF. Due to the fact that the server must maintain a permanent connection to the Internet without user interaction, we implemented it as an Android service, i.e., a background process.
One important issue appears when one would like to get data from a device because the URI used to query the content providers has a local meaning. In the outside world, 1 RDF Calendar vocabulary: http://www.w3.org/TR/rdfcal/. 2 Geo location vocabulary: http://www.w3.org/2003/01/geo/ .
Fig. 2. An example of using the RDF Browser. the URI used to query the address book of two different persons will be the same, but the content of the address book will be different.
The server principles are quite simple. In the beginning, the server receives a request from the outside. Then, it dereferences the requested URI, i.e., it translates the external URI into an internal one, which has meaning inside the Android platform. The RDF Server sends it further to the RDFContentResolver. In a manner similar to the one explained for the RDF Browser the set of triples is obtained but, before sending this set to the server, the URIs of the triples are externalized and the graph is serialized using a port of Jena under the Android platform.
The URI externalization process translates the local URI content://authority/path/to/data into the dereferenceable one http://deviceIPAddress:port/authority/path/to/data. Reversing the translation of such a URI is possible since both the authority and the path are kept during the externalization process.
Usually, mobile devices do not have a permanent IP address and thus, the externalized URIs are not stable. To overcome this, a dynamic DNS client34 can be used.
In addition, the server supports a minimal content negotiation mechanism. If one wants to receive the data in RDF/XML, it will set the MIME types of the Accept-type header of its request to "application/rdf+xml" or to "application/*". In the opposite case or when the client sets the MIME type to "text/plain", the data will be transmitted in an N-TRIPLE format. Not only the requester has the opportunity to express its preferences regarding the format of the received data, but the default format of the transmitted data can be specified in the server settings, as well the port on which the server can listen on and the domain name server for it. 3 Dynamic DNS Client: https://market.android.com/details?id=org.l6n.
dyndns&hl=en. 4 DynDNS: http://dyn.com/dns/ .
Fig. 3. RDF Server response.
An example can be found in Figure 3. In this scenario, the user retrieves information about the fourth contact from the device address book. The request is processed by the RDF Server in a manner similar to the one of the RDF Browser. 6
Technical Details
The RDF Server included in our architecture eases the access of the user to the RDF
data found on the web. For that purpose, we wanted to reuse an existing semantic web
framework, such as Jena or Sesame. Yet they are not suitable to be employed under
the Android platform (the code depends on some libraries that are unavailable under
Android). There are a few ports of these frameworks to Android: Microjena5 and
Androjena6 are ports of Jena and there exists a port of Sesame to the Android platform
mentioned in [
A problem that arises when we use this framework is that the size of the application increases substantially. This problem could have been avoided by reimplementing only the Jena modules that are needed in our architecture. Still, we would like to improve our architecture by adding more features (such as a SPARQL query engine) that require additional modules to those used to read/parse/write RDF, available in Jena.
A tool that we found useful in our development process was ProGuard. ProGuard7 is a code shrinker, optimizer, and obfuscator. It removes the unused classes, methods or variables, performs some byte-code optimizations and obfuscates the code. The tool proved to be efficient in reducing the size of our application (our framework including Androjena) by half, i.e., its initial size was 6.48MB, and after we applied the tool it diminished up to 2.98MB.
The existence of such tools as ProGuard, is a step forward in the continuous battle between applications that require a considerable amount of space for storing their code and devices with a reduced memory storage.
We are currently examining how to query the device data using SPARQL. There are
two main ways of doing this:
5 http://poseidon.ws.dei.polimi.it/ca/?page_id=59.
6 http://code.google.com/p/androjena/.
7 http://proguard.sourceforge.net/.
– creating a new RDF content provider which relies on a triple store to deposit the
data [
At the moment, we are investigating the second option. There are several available tools that can make the translation from SPARQL to SQL, like Virtuoso or D2RQ. However, these tools solve only half of the problem because the SQL queries have to be adapted to the ContentProvider interface, i.e., the queries have a particular format, different than the SQL one. This interface allows for querying only one view of a specified table at a time, hence it is not possible to ask Content Providers to perform joins.
Further challenges regarding security must be taken into account. The user of the application should be able to grant or to deny the access to its personal data. A specific vocabulary should be used in order to achieve this 8. More that that, the dangers of granting system access to a third-party user can be avoided by using a secure authentication protocol 9.
As can be seen, there are still technical problems in implementing a full RDF framework at the core of Android. Specific solutions must be developed. 7
Conclusion Involving Android devices in the semantic web, both as consumers and providers of data, is an interesting challenge. As mentioned, it faces the issues of size of applications and URI dereferencing in mobility situations.
A next step is to provide a more fine grained and structured access to data through SPARQL querying. This promises to raise the issue of computation, and thus energy, cost on mobile platform.
A further issue will be the control of privacy in such a framework. But here too, we think that semantic technologies can help. 8 http://www.w3.org/wiki/WebAccessControl. 9 http://www.w3.org/wiki/Foaf+ssl. The Web of Radios - Introducing African Community
Radio as an interface to the Web of Data Anna Bon1, Victor de Boer1, Pieter De Leenheer1, Chris van Aart1, Nana Baah Gyan1, Max Froumentin2, Stephane Boyera2, Mary Allen3, Hans Akkermans1 Abstract. The World Wide Web as it is currently deployed can only be accessed using modern client devices and graphical interfaces, within an infrastructure compassing datacenters and reliable, high speed Internet connections. However, in many regions in developing countries these conditions are absent. Many people living in remote rural areas in developing countries will not be able to use the Web, unless they can produce and consume voice-based content using alternative interfaces such as (2G) mobile phone, and radio. In this paper we introduce a radio platform, based on a use case and requirements analysis of community radio stations in Mali. The voice-based content of this radio platform will be made publicly available, using Linked Data principles, and will be ready for unexpected re-use. It will help to bring the benefits of the Web to people who are out of reach of computers and the Internet. 1
Introduction The World Wide Web is perfectly adapted for use by people in developed countries. It is visual, text-based, and mainly written in English or a few other world languages 4. The Web depends on the availability of computers, datacenters, glass fiber backbones, fixed and wireless networks, 3G mobile telephony and transport of large volumes of data at high speed. In remote rural areas in many developing countries, conditions are different. Poor infrastructure, lack of equipment, low levels of literacy, and use of under-resourced local languages, seriously hamper the access to the Web for many people.
There is a general consensus that the global Information Society must benefit all people
in the world. The United Nations Millennium Declaration contains a commitment for
developing a people-centred, inclusive and development-oriented Information Society
4 Wikipedia, http://en.wikipedia.org/wiki/Global Internet usage,Global Internet Usage
so that people everywhere can create, access, utilize and share information and
knowledge to attain the internationally agreed development goals and objectives, including
the Millennium Development Goals.5
Yet, in many rural regions in Africa community radio is the only source of information.
People have radios at home and listen to programs broadcast in local languages every
day. Many people have access to simple voice-based (2G) mobile phone, but text
messaging is hardly used [
The availability of both mobile phone and radio is opening opportunities for new services. E.g. radio listeners phone to the radio station and leave voice messages that they want to have broadcast, or react to popular radio programs leaving news, opinion, regional information etc. Community radio here operates as an important local information hub, where people bring information for further dissemination.
Radio stations in rural areas in Africa operate under harsh conditions. Only the largest and state financed radio stations have a computer and an internet connection. Due to lack of funds many radio stations still use old-fashioned, analogue equipment, such as tape recorders. Yet, it is in the line of expectation that more and more radio stations will have computers and an internet connection in the coming years.
In the current situation the information broadcast by the community radio is volatile: it is not stored and kept for later access or re-use. Radios do not have means to manage, reuse and index this voice-based content.
In this paper we introduce a radio platform as a new interface to the Web. It enables
management of radio content in an efficient way, making it accessible and searchable,
so that it can serve a broad audience, e.g. Africans in the diaspora, who want to have
news from their home villages6
Additional, the voice-based radio content on this radio platform might be linked to other
data sources on the Web, enabling community radios in Africa to become an interface
to the Web of Data. An example of a system that manages market information based on
Linked Data principles and produces voice-output as broadcasts for African community
radios, is described by De Boer et al. [
Contributions of this paper are: – A radio platform with both a web and a voice-based mobile interface that allows content creation, retrieval and indexing of spoken radio content. 5 UNMD, United Nations Millenium Declaration, General Assembly resolution 55/2. United
Nations, New York, 2000 6 Communication possibilities with people living in the diaspora, are described by Serigne Mansor Tall in: Les e´migre´s se´ne´galais et les nouvelles technologies de l’information et de la communication. http://www.unrisd.org – African community radio, introduced as a new interface to the Web of Data
This paper is structured as follows. In section 2 we describe related work. In section 3 the architecture of the radio platform is described, the use cases collected from three different radios in Mali, as well as the principles used to manage the content. In section 4 we describe the challenges related to the organization of the voice-based radio content. In section 5 we discuss future work that must be done on the Web of Radios, including the sustainability aspects. 2
Related Work Related work on the development of a similar platform was done in the Freedom Fone7. Freedom Fone is a project initiated by The Kubatana Trust of Zimbabwe, a civil and information activist platform from Zimbabwe. Freedom Fone is open source software for creating audio content using phone. Freedom Fone provides a voice platform similar to the basic setup proposed in this paper, but without the Linked Data enabled data management.
Research on speech recognition started in the 1930s and resulted in commercial
deployments of voice-based services in the 1970s. Major achievements on language
recognition, mainly for English, took place in the 1980s and 1990s and culminated in the
development of VoiceXML by the W3C Voice Browser group, in 1999, facilitating and
standardizing the development of voice applications [
Sheetal Agarwal et al. from IBM Research India, developed a system to enable
authorship of voice content for 2G phone in a web space, they named the WWTW or
World-Wide Telecom Web. The system is not connected to the Web, therefore not
allowing access by third party search engines. The system represents closed web space,
within the phone network. Especially the lack of open search possibility constrains its
growth [
From Burundi a system has been reported [
A related project on the Semantic XO and Linked Data for developing countries is
described by Gue´ret et al. [
De Boer et al. [
The radio platform The design of the radio platform described in this section is based on extensive use case and requirements analysis, performed in Mali, with the collaboration of radio journalists from community radio stations. The research was done as part of the Foroba Blon9 project10, funded by the International Press Institute, and the VOICES project11, partially funded by the EU, within the 7th Framework Programme. The Foroba Blon project is aimed at supporting and promoting citizen journalism in developing countries. The VOICES project is aimed at developing innovative mobile voice services to support users in underprivileged communities in African countries.
phone users Fig. 1. Conceptual design of the radio platform as a voice-interface to the Web for people who are out of reach of computers and the Internet, but do have phone or radio.
Hard Disk Space Battery Power ts Memory n i tra Network Bandwidth s n Co Available Connectivity Reaction Time Processing Bandwidth
Ranking Measures Similarity Frequency Size
Frequency
Fig. 1: Ranking Measures vs. Constraints 3
We discuss two very different case studies: the SemanticXO and visualization of
census data, and the way rankings and resources bounds relate in them.
An environment with obvious constraints is given in the SemanticXO project.
The XO laptop is part of the One Laptop per Child (OLPC) project. The aim
of OLPC is to create educational opportunities for the worlds poorest children
by providing each child with a ‘rugged, low-cost, low-power, connected laptop’.
The SemanticXO is a project which aims ”to provide an infrastructure that is
needed to integrate semantic information from the Web of Data into programs
(...) running on an XO computer” [
Software Stack Currently, the SemanticXOs software stack contains [
SemanticXO, or elsewhere on the Web of Data.
The SemanticXO uses this software stack to store information as a Data Graph in the triple store. After the graph is stored in the triple store, it is available to other XOs via the HTTP Server or the API.
Problem description XO’s are often used in a network with similar XO’s, where a school server functions as an access point with internet connection. Currently, distributing information within such difficult. Moving information from one XO to another requires both XOs to be on-line, and run the same program (called ‘activity’). Situations where the recipient of information is offline, where information may be shared asynchronously, or where information needs to be shared regardless of any running activity, are currently impossible to deal with.
The SemanticXO offers a framework with which to approach this problem. Currently, objects containing meta-information are stored locally as a data store object (DS-Object). These objects contain meta-information, and contain (a combination of) attribute-value pairs. More complex information such as images are stored separately. Generic shipping of objects requires a DS-Object to be wrapped as a Semantic DS-Object (SDS-Object), which is stored as a graph in the triple store. Via the HTTP server or the API, this information is automatically available to other XO’s in the network. Using this infrastructure, the nodes can share and sync objects. This creates a network of loosely coupled triple stores between which information has to be distributed.
Constraints & Ranking We will discuss the constraints and rankings in this case study, using the claims from the previous section.
(1.) The world is resource bound Internet connections are unreliable or have a very limited bandwidth, and hardware may not be on-par with regular laptops and desktops, to name but a few.
(2.) Deal with these constraints Not dealing with constraints such as network bandwidth will most likely result in an inadequate distribution of objects. Not all objects may be distributed to all nodes, which means SDS-Objects run a risk of not being delivered to the intended recipient.
(3.) We need ranking Constraints may limit the number of SDS-Objects to be shared, which means a subsection of the SDS-Objects should be synced. This is essentially a top-k ranking problem, where only the most useful objects are cached and synced, and others are ignored.
(4.) Ranking: it’s all in the data Relevant ranking measures are Recency (i.e. how long ago is this graph created or modified), Size (i.e. how big (in bytes) is the graph), and Frequency (i.e. how often is this graph synced)
(5.) We need explicit bounds The resource bounds of the SemanticXO are
either caused by the hardware specification[
(6.) Tell me your bounds and ranking, and I deal with them By combining the resource bounds (5), the rankings (4) and the links between both (fig 1), the application can deal and adapt to its resource bounds. We consider the following relations between the constraints and these rankings measures: 1. Hard Disk Space: With limited disk space available, giving precedence to smaller (size) objects will allow the SemanticXO to store a larger quantity of objects. 2. Battery Power: When battery power is running low, giving precedence to recency and frequency will make sure the old objects which have not been synced that often are ordered higher than others, because older objects in the queue might indicate these objects are not spread through the network. 3. Available Memory: Processing large graphs might require more memory.
When available memory is low, precedence should be given to smaller graphs.
Fig. 2: Census Visualization 4. Network Bandwidth: A low network bandwidth should decrease the importance of the size of SDS-Objects. This way, at least the smaller objects will still be distributed via this node. 5. Available Connectivity: When the SemanticXO is connected to a very limited number of other nodes, the graphs which are not very well distributed in the network should be given precedence. This means objects which are not often synced (frequency ) and older object (recency ) are ranked higher than others.
This framework is particularly useful for the SemanticXO, because hardware may differ between XO’s. An XO server has completely different hardware (bounds) than a regular XO laptop. Using the described framework, the server and laptop can both optimize their performance based on their own resource bounds. 3.2 Visualizing Census Data To show that resource bounds are not limited to the environment of developing countries we describe a totally different application where Dutch historical census data is visualized. This use case is part of the Data2Semantics project2, which focusses on the problems of ‘how to share, publish, access, analyse, interpret and reuse scientific data’.
Visualizing Census Data This dataset contains Dutch census counts from between 1795 and 1971, and covers demographic information, such as gender, age, location, marital status, occupation, household and religion. An example of an application making use of the (convert to RDF) dataset is shown in figure 2. This visualization has several selectors: 1. Variable selection: The variable of which the distribution is shown on the map 2. Time: The year of the census count 2 www.data2semantics.org 3. Zoom: Zoom in or out of the map 4. Age: Age range for which to visualize the data for 5. Aggregation Level: The aggregation level for the visualization. Changing this value changes the granularity of the colors on the map.
Because this visualization makes use of a SPARQL endpoint, these selectors correspond to SPARQL queries. In this respect, we can consider the variable selection, time, and age as filters, and the aggregation level as a ‘group by’. Constraints & Ranking We will discuss the constraints and rankings in this case study, using the claims from section 2.
(1.) The world is resource bound This visualization application is clearly resource bound. Users will expect near to any-time responses when changing the sliders. Additionally, the application will have to deal with current (consumer) hardware and network constraints.
(2.) Deal with these constraints Not adapting to the constraints leads to situations where the response time (when changing a slider) is too long.
(3.) We need ranking A way to deal with these constraints is by pre-executing queries. Executing all possible setting combination will not be feasible, however: there are too many possible combinations of queries. Deciding which queries to pre-execute, though, is a ranking problem/
(4.) Ranking: it’s all in the data Relevant ranking measures are: 1. Query similarity: Difference in similarity between current view, and the possible query. The bigger the difference, the wider the ‘window’ becomes which is being pre-executed. 2. Frequency of usage: Selectors often used are more interesting to pre-execute.
Therefore, queries where the filter value of a frequently used selector is different from the filter value in the SPARQL query of the current view, should have precedence. 3. Size of expected number of results for a given query: e.g., decreasing the aggregation level increases the retrieved number of triples.
(5.) We need explicit bounds The resource bounds of this visualization involve: available memory, network bandwidth, reaction time of the user, and processing bandwidth of the user.
(6.) Tell me your bounds and ranking, and I deal with them By combining the resource bounds (5), the rankings (4) and the links between both (fig 1), the application can deal and adapt to its resource bounds. We consider the following relations between these constraints and the rankings measures: 1. Memory: Limited memory implies limited available space to store the query results. This should give precedence to queries where the expected query result is low. 2. Network bandwidth: The smaller the network bandwidth is, the smaller the expected size of query results should be. 3. Reaction time: More time might allow for a smaller query similarity, which results in a bigger window size for the pre-executed queries. A larger interval allow for a larger query result, as there is more time available for the response to be generated and processed. 4. Processing bandwidth: The more information a user is able to process, the larger the size of the query result may be, and the bigger the allowed windows size may be (query similarity ). 4
This paper showed a perspective with which to consider applications in a (dynamic) resource bound environment. The presented framework should allow semantic web applications to adapt to changing constraints, by making use of the explicit and intrinsic orderings and rankings in the data. We showed the generic link between ranking and constraints, and put these into context of two very different case studies. This framework is not complete though. A roadmap towards completion requires discussion on some unanswered questions, and an implementation/evaluation of this framework. Current open questions are: – The ranking measures mentioned in section2 are very broad: recency, size, frequency and similarity. Can we keep these notions on such a generic level? Or are these ranking measures open to interpretation, depending on the application domain? – What are the generic relations between constraints and ranking measures? For both use cases we show some links. Are these generic? And does dealing with one constraints influence other constraints? Is there a cost in dealing with constraints. – How do the ranking measures relate to each other in such a dynamic system?
How is the final ranking score determined? – How to define interaction constraints? Different users interact differently with a system. – How does this approach hold up against other methods such as the use of access control, differentiated services, or formal maximization under constraint? The main question seems to be: does explicating these concepts and relations work in practice? This requires further implementation and evaluation of this framework in one of the use cases.
Acknowledgements. This publication was supported by the Dutch national program COMMIT.
IKM Emergent m.powell@pop3.poptel.org.uk Abstract. This paper looks at the purpose of developing linked open data applications of value to poorer and less well connected potential users and considers some of the challenges that need to be met if this purpose is to be achieved. It stresses the different societal contexts in which such applications will be deployed and describes some of the potential negative and unintended consequences of enthusiastic but ill prepared initiatives. It describes the development information environment as one of overlapping information ecologies, which, as the metaphor implies, contains both interdependencies and examples of predatory, counter developmental behaviours. From this analysis it suggests a few key areas which combine the potential of real value for users with exciting and ground breaking technological challenges. 1 The Downscale 2012 workshop asks software producers1 to think of the four billion people who do not have access to the internet when designing linked open data platforms and to aim to reduce rather than add to the digital divide. This request seems entirely consistent with the purpose behind many existing linked open data initiatives, which seek to promote greater equity in governance through enabling higher levels of transparency and accountability. It could indeed be said that it is hard to see the point of an interest in linked open data solutions without having an interest in their potential social impact. If this is indeed a common perspective within the linked open data community, then the questions of what social impact is intended and how it may be achieved need to be taken seriously.
This paper, whilst commending the Downscale initiative, argues that the relationship between information and poverty is far more complex than a simple lack of access to communications channels or indeed to information itself. It suggests that deeper study of the potential of information to combat poverty is required. Such deeper study needs, in each situation, to include the specific dynamics of any particular type of information and of the situation in which any 1 To avoid confusion, we use the word “development” to relate to the field of international development, struggles against poverty etc. and thus must use another word to describe the development and developers of software and other technical solutions intended user group finds itself. However, some general points can perhaps be outlined here. The desired result of such preparatory work is to identify issues in people’s informational needs which pose new technical challenges as well as to become aware of and build meaningful collaborations within the social and organisational contexts in which any technical solutions will be applied. 2
This paper is also located within a discourse about “international development”. The author participated at a panel discussion at ICTD 2011 at which considerable scepticism was expressed by a predominantly technical audience about the relevance of the international development sector. Some argued that the idea of top down investment in processes of poverty alleviation involved outsiders telling other people what they should do and was therefore patronising. Others suggested that such efforts had been largely ineffective and therefore need not be taken into account. In this author’s opinion, there is some merit in both arguments. It is also undeniably the case that much economic and social development, including innovation with ICT, emerges from the internal dynamics of local societies and has no relation to the purposeful efforts of international development actors. That said, the sector is a significant presence in many parts of the developing world. Definitions of what constitutes “official development assistance” vary, but the OECD2 calculated that it amounted to USD 127.6 billion in 2010. When we consider that some USD 2-3 billion of that amount is spent by professional development organisations just on their own information systems, including their linked open data platforms, we can see that the sector is inevitably going to impact on the information landscape. It has the potential both to offer resources and co-ordination to solutions which will help the poor or, as has been argued elsewhere, to worsen the digital divide3. It is also the case that history of international development efforts is littered with examples of “technical experts”, including in the ICT4D field, failing to recognise or adapt to the socio-cultural or physical specificities of the new environments in which they were working and thus failing spectacularly. Thus for the lessons of its failures as much as its successes, as well as for its exchanges of ideas and the opportunities it offers to build collaborations, it would be unwise to ignore the work of the development sector.
Indeed one starting point for any information and development initiative
should be a clear understanding of what is meant by development. Leaving aside
the large number of examples which could legitimately be given of development
assistance being applied for the direct economic or political benefit of the donor,
2 OECD, Query Wizard for International Development Statistics, accessed September
15th 2011
3 See Powell, Davies and Taylor, 2012 “ICT for or against development? An
introduction to the ongoing case of Web 3”. IKM Emergent Working Paper, in press. The
paper takes an historical overview of the use of ICT within the devlopment sector
and looks more particularly at early linked open data initiatives within it.
there are longstanding and legitimate debates on what types of development
strategy are likely to be most effective. Is it more useful to attempt to support the
general development of a poorer country in the expectation that resulting growth
and educational improvements will spread to all sectors of society and, because
of their wider base, be more sustainable? Or is it better to target interventions
at the specific needs and challenges of the poorest and most marginalised? As
Wright et al [
This question is fundamental for any information-led development initiative, including those concerned only with data. If “development” initiatives are intended to be of direct benefit to the poorest, this implies improving the relative position of the poor in relation to the rest of society, or, if we are thinking in terms of societies, the relative position of a poor country in relation to global levels. Given that information by itself is of little value without the capacity to make use of it and that such capacity is strongly linked to educational levels, it follows that simply making information more generally accessible is likely to worsen not improve such relative positions. Wright et al., in the paper cited above, highlight the danger of what they call “elite capture of transparency”4, whereby information coming from or intended to benefit poor people is captured and exploited by others, a process which would be familiar to students of the uptake and use of public health services in Europe. Thus simply enabling the provision of data or information alone is not adequate. Questions need to be posed as to what information would be of particular value to particular groups.
Equally important are issues of how information is provided and curated and to what extent the people using and producing the information have any control over the process. These issues relate to power and profit but also to culture. All societies are based on sets of cultural understandings which may be more or less homogeneous, more or less dynamic in each situation. In recent years changes in informational behaviours have led to new understandings of the interaction of information and culture in many parts of the world. However, as an UNRISD 4 A full case study, also referred to by Wright, is Benjamin, S. Bhuvaneswari, R.
Rajan, P. and Manjunatha 2007 “Bhoomi: ’E-Governance’, Or, An Anti-Politics
Machine Necessary to Globalize Bangalore?” A CASUM-m Working Paper http:
//casumm.files.wordpress.com/2008/09/bhoomi-e-governance.pdf
workshop in the run up to the first global summit on the “Information Society”
(WSIS) concluded: “it is a serious mistake to assume that they constitute a
uniform process globally or share a common destination, rather than a variety
of new processes each influencing and being influenced by the society in which
they are taking place.” [
One feature of the power imbalances in the development field is the influence of large, well resourced institutions using large data sets from which to generate generalised policy approaches which are then applied to local situations. This process can ignore the specificities of local circumstances and limit the freedom of people working at more local levels to develop policy which seems appropriate to them. In a similar vein, such organisations (and many others) justify their roles by making extravagant claims about the value of their knowledge and 5 In Chile and ogd-feasibility/
Ghana.
See http://www.webfoundation.org/projects/ implying that the main challenge of development is getting this knowledge to the people who need it. This was an explicit argument behind the World Bank’s claim to be a “knowledge bank” in the late 1990s. Such attitudes lead to one-way information flows which are often not based on sufficient information about their intended recipients for the information to be of much value. As importantly, it can be hard for information about the realities on the ground, those realities which the whole effort is intended to improve, to get fed into and used in these policy making and communications processes.
Another aspect of the development information ecology is that although there is a certain attempt to maintain a common purpose of “co-operation for the common good” across the sector, most development organisations are in an increasingly competitive environment when it comes to seeking funds. This means both that they try to emphasise the origin and value of “their” information, for example by seeking to encourage direct traffic through their own web site and also that they often have few resources to invest in collaborative information initiatives. This, as well as the complexity of the quantities and ranges of information that is relevant to development, has resulted in a very fragmented information environment. With a few exceptions, mainly limited to certain well defined communities of interest, information resources are highly dispersed, metadata is poor, and the inbuilt algorithms of most search engines tend to recognise and reproduce the influence of the more powerful information providers.
If the above concentrates on the information ecology as it exists for
professional development organisations working at international level, the situation
for community level organisations and for individuals, marginalised or living in
poverty is even worse. A fragmented information environment is inconvenient
and inefficient for someone working in an office in Geneva, for someone relying
on a pay per minute connection in an internet caf´e in Africa it is impossible.
Despite all the rhetoric about knowledge equating to power and the large ICT
and communications budgets of the sector as a whole, there are incredibly few
resources made available to support information processes which go beyond the
idea of the passive recipient. “Passive recipients” are of no value to development.
What is needed is for people to respond to information, understand it, adapt it
to their own circumstances and to use it: or, in other words, to go through the
process whereby information becomes knowledge. As has been argued elsewhere:
Formal education undoubtedly helps, but so can many other forms of
human interaction. People need to be able to validate information and
to think through if and how any of it may be useful to them. In this
context, and whatever other mechanisms may be available to help the
process, connections with other people are essential, not only as sources
of information, but also as means of validation, reflection and action. This
is true for everyone - from the fraternity clubs of elite US universities to
networks of the most poor and marginalised. “Ki rafl´e du ki amul yeer´e
way´e moy ki amul nit”, as a Senegalese proverb has it, “the poor person
is not the one without clothes but the one without anyone.” [
The same chapter argues that most people, however marginalised, take part in a number of informational spaces, “some geared to family matters and social obligations, others related to work and income, some perhaps related to politics and governance and to faith” (p 134). These spaces, each of which will have their own rules and norms depending on their character and purpose, are seldom seen as part of the development knowledge ecology, but in fact they are key to knowledge being created and used by the people who’s actions determine whether or not any development actually takes place. At least some of these spaces are also open to interaction with external supporters but very little work has been done at this level to really understand what types of information are most useful to participants, to what extent and how digital platforms can be used, what issues of privacy and restricting access arise. It is an area of huge potential but also of many difficulties.
Value: in many countries much government data is inaccurate or incomplete, often out of date and, whatever the legal requirements may be, often hard to hold of6. In the absence of reliable free information, some have sought to develop enterprise models which either reward the generation of data and/or seek to charge for its use7. These models may offer better value than public information but, if the end-user has to pay for the data thus provided, are likely to be of less benefit to the very poor.
Security: in the wrong circumstances even the most apparently innocuous
information can be misused. Ushahidi8, in Kenya, rightly attracted a lot of
attention by its ability to use crowd-sourcing to map emerging troublespots
in the violence that erupted after that country’s elections in 2008. In doing so,
the organisation, locally based and in touch with the many of the elements
in Kenyan society working to end the violence, will have had to exercise
judgement as to what information to make public. The same information,
used differently, could have had disastrous consequences as was demonstrated
by the use of Radio Mille Collines as an agent of the genocide in Rwanda.
Cost: the factor of cost is ever present. Amidst the marvelling at the growth of
mobile phone usage in Africa, issues of inclusion and exclusion can easily get
lost. As mobile phones become the main vehicle for communication across
extended families, the poor are left with the choice of paying up - sometimes
over 20% of their income - or removing themselves from their most life
affirming community. Likewise, and in a cautionary tale for any proposed support
of informational spaces, poor women in Zambia found themselves excluded
from networks specifically set up to “empower” them (See, for example, [
All of the above pose a number of challenges for anyone wanting to develop linked information solutions in a way which goes beyond the basic connection to actually the poor and the marginalised to improve their conditions. It underlines the need for real understanding of local contexts, the desirability of working with local partners, and the value of participatory engagement with information providers and users if good choices are to be made about process, content and platform. Some of the challenges are of a general nature, others are more directly connected to issues of linking data and information.
– The need for caution in becoming over reliant the data of powerful institutions like the World Bank and OECD. It is hard to avoid using their data and it is of course positive that they are increasingly making their data openly available, so that it becomes a valuable resource to query. On the other hand, constant reference to these sources can appear to reinforce their dominant roles in the development information ecology and privilege the types of evidence favoured by global policy makers over evidence which enables learning from local contexts. – In the same vein, there is a pressing need to make sure all relevant voices are heard and that data and other information from the grass roots becomes more visible – The term “linked open information” rather than “data” is used in the subheading above because, if the aim is to support the information needs of the poor, there is no point in privileging data over other types of information. In the situations we are discussing, people are generally lacking many types of information. This poses the challenge, which to some extent has been explored in the IKM Emergent Programme10, of using RDF and other datalinking tools to create links to metadata about other forms of information and to related social media discussions 9 Plenary floor exchange at First Global Congress on Community Networking,
Barcelona, Nov 2000 10 See http://wiki.ikmemergent.net/index.php/Workspaces:1._Information_ artefacts and http://wiki.ikmemergent.net/index.php/Workspaces:1: Linked_Open_Data – Work on data itself, especially government data, in developing countries is likely to involve political and technical issues to improve its availability, accuracy and timeliness as much as technical development. It may also involve initiatives aimed at deriving data from crowd sourcing, a practice of which there are already examples from India and Kenya. – The issues of the platform and the affordability of its use is crucial. Mobile Phones are an obvious choice but the cost of their use, particularly for higher bandwidth applications can be prohibitive. It may make sense to explore the potential for partnerships with resource centres, telecentres or, where they exist, libraries. – Reusing schema can generate links simply and efficiently but again raises issues of politics and influence in whose schema are used. Standards also need to be established to avoid the development of an “official” development linked open data in which the powerful professional organisations exchange data and which excludes interaction with other sources of information and data – Developing thesauri and ontologies capable of transcending the multi-lingual, multi-disciplinary realities of development will be a real challenge for what are referred to in the literature as “emergent ontologies”, “heterogeneous ontologies” or “dynamic networked ontologies”. The issue of “semantic interoperability” is also highlighted in the “Report on Open Government Data in India” cited above. 5 Finally, as the authors of a report on women’s use of ICT in Mozambique conclude, there are limits to what can be achieved by the use of ICTs alone. Their impact can be immeasurably strengthened if they can be deployed alongside other more traditional tools of empowerment.
“Literacy is key - without literacy there can be no empowerment,
particularly for women and girls. We therefore strongly recommend the
improvement of women’s literacy in rural areas. We believe that women’s
literacy, combined with increased relevance of content, could result in
computer-related ICT tools becoming an asset to women’s pursuit of
the means for survival and for control of their lives.” [
What this conclusion suggests, apart from the point it so clearly makes, is that the process of linking information with meaningful change is far from straightforward and is unlikely to be taken very far by the development of one set of technologies, produced in isolation. Collaboration in more broadly based change efforts perhaps offers a better chance of greater impact as well as providing an opportunity for the discussions of assumptions, needs and options with people, rooted in their communities, to better understand their needs and create a process of mutual learning.
Philippe Cudr´e-Mauroux1, Gianluca Demartini1, Iliya Enchev1,
Christophe Gu´eret2, and Benoit Perroud3? 1 eXascale Infolab, University of Fribourg, Switzerland
{firstname.lastname}@unifr.ch 2 Vrije Universiteit Amsterdam, the Netherlands
c.d.m.gueret@vu.nl 3 VeriSign Inc., Fribourg, Switzerland
bperroud@verisign.com Abstract. Web of Objects and Linked Data applications often assume that connectivity to data repositories and entity resolution services are always available. This may not be a valid assumption in many cases. Indeed, there are about 4.5 billion people in the world who have no or limited Web access. Many data-driven applications may have a critical impact on the life of those people, but are inaccessible to those populations due to the architecture of today’s data registries. In this paper, we point out the limitations of current entity registries when deployed in poorly connected or ad-hoc environments. We then sketch new architectures based on IPV6, structured P2P networks and data replication for entity registries that could run in ad-hoc environments with limited Internet connectivity. 1 Data registries are critical components of the Web architecture and are widely used in every-day web activities. For example, domain name registries are databases containing registered Internet domain names. They are necessary for all web users wishing to visit a website knowing its URL (e.g., hostname) rather than its IP address. Thanks to the Domain Name System (DNS) infrastructure, such information can be obtained by recursively resolving a domain name to an IP address.
Another example of registry is the Digital Object Architecture (DOA, see Section 2). It allows to assign unique identifiers to digital objects (e.g., scientific publications) which can then be univocally accessed by users. Their identity will thus last in time even if their physical location may change (similarly to the IP address associated to a domain name, which often changes over time).
In situations where data registries are not continuously accessible, the user experience can be strongly limited. As a basic example, if the DNS server used ? Authors are listed in alphabetical order. by a client computer is not connected to the rest of the DNS hierarchy, then a very restricted set of Internet domains can be resolved to their IP addresses.
In addition to those traditional registries offering hash-table like functionalities, online infrastructures and applications are increasingly turning to more flexible registries containing information about general objects or entities (i.e., object registries and entity registries ) to power data-driven applications. Emerging examples of that trend are DBPedia4 and Freebase5 which, given an entity identifier (e.g., a URI), provide semi-structured metadata about the entity. Another example are the registries used for the Web of Objects, which mediate information between networks of digital devices connecting to each other, enabling information publication or integration in sensor networks or smart building contexts for example. In those cases, the infrastructure needed is more complex than a traditional “Hostname-IP” DNS system and is closer to a global registry mapping unique identifiers to arbitrary structured data.
In situations where such registries are not continuously accessible, however, the user experience can be strongly limited. As a basic example, if the DNS server used by a client computer is not connected to the rest of the DNS hierarchy, then a very restricted set of Internet domains can be resolved to their IP addresses. In an object registry context, discontinued access to the registry typically results in the impossibility to publish data or issue object queries.
In this paper, we argue that data registries increasingly represent an essential part of today’s Internet ecology (see Section 2 for a few examples), but that their current architecture precludes their use in many important contexts. For example, we can envision ad-hoc environment where the nodes self-organize without having access to third-party registries. In such transient and poorlyconnected environments, nodes have a clear need to discover, connect, and exchange (structured) information with related entities locally and should be able to do so without resorting to any outside registry. Another interesting context is data-intensive object applications, where nodes have to discover and exchange data about very large numbers of entities and should be able to do so in a peer-to-peer manner whenever possible.
We propose in this paper different approaches to overcome the limitations of existing registry solutions for poorly connected environment or localized environments. First, we point out the limitations of current registry architectures, before proposing solutions that take into account the limited connectivity of the peers and enable the management of digital information in ad-hoc environments.
The remainder of this paper is structured as follows. In the rest of this section we motivate our work: we show why it is important to consider entity registries for ad-hoc environments and which are their benefits. In Section 2 we briefly describe existing architectures for data and entity registries. Section 3 highlights the problems of current solutions when applied to poorly connected or ad-hoc environments. Section 4 presents two alternative solutions to exploit entity registries based on IPV6 addresses and on P2P networks respectively. Finally, we conclude and discuss future work in Section 5. 4 http://dbpedia.org/ 5 http://www.freebase.com/
Use case: Internet-less Mesh Networks
A rapidly increasing number of applications—such as open social applications,
applications relying on governmental data (data.gov) or Linked Open Data6—
assume an ubiquitous and continuous access to the Internet in order to power
data sharing and data-driven applications. As pointed out in our previous work
[
We can for example imagine a XO laptop connected to a server (e.g., at school) downloading some Web pages or Wikipedia articles. Afterwards, the laptop is moved to a different location with no Internet connectivity but with the possibility to connect to other XO laptops. This scenario enables the sharing of previously downloaded documents with others who had no possibility to obtain them from the Internet, and the local publication of new entities.
In this context, it is often important to identify entities in all documents to enable entity-centric document aggregation, semantic of faceted search. Such aggregations may, for example, support learning applications that present the set of documents users should read when they want to learn about a specific entity (e.g., “Malaria”).
Extraction entities from HTML text may be performed automatically by
tools running on the XO laptops or even manually exploiting crowdsourcing,
which can address the problem of limited computational resources available [
DNS The Domain Name System (DNS) is the system used on the Internet to resolve domain names to their corresponding IP addresses. Domain resolution works in a hierarchical manner; the top of the domain name space is served by so-called root name servers, pointing to authoritative name servers maintaining authoritative information for the top-level domains (a.k.a. “TLDs”, such as “.ch” or “.com”). The authoritative name servers responsible for the TLDs point in turn to further name servers, responsible for second-level domains (e.g., “unifr.ch”), and so on and so forth to process each domain name label iteratively until the last iteration, which return the IP address of the domain name queried. In practice, domain names are often cached at various levels, for instance at the client-side, or at the level of the DNS server provided by the Internet Service Provider in order to limit the load on authoritative DNS servers.
Though originally not designed for this purpose, it is be possible to extend
the current DNS infrastructure to create a full-fledged entity registry. In that
context, we recently suggested an extension of the DNS [
DOA The aim of the Digital Object Architecture (DOA)8 is the management of digital entities over potentially very long timeframes. There are three distinct components in DOA: – the Resolution System (Handle System) – the Digital Object Repository (DORepository) – the Digital Object Registry (DORegistry) The principal function of the Handle system is to map known identifiers into handle records, containing useful information about the digital object being identified (e.g., IP address, public key, URL etc.). Every identifier has two parts: a naming authority (or prefix) and a unique local name under the naming authority suffix, separated by “/” (e.g. “10.1045/january99-bearman”).
The collection of all local names defined under a certain prefix defines the local handle namespace under that prefix (something similar to a root zone in the case of DNS). All the local namespaces (i.e., all prefixes) define the handle namespace and a prefix can be considered as a top level domain. More namespaces for Local Handle Services (LHS) can be defined in a hierarchical fashion 8 http://www.cnri.reston.va.us/doa.html under the Global Handle Registry (GHR), thus the Handle system provides a hierarchical service model.
The DORegistry provides services like browsing, searching, repository and federation for collections of digital objects that can be distributed across multiple sites including other DO Registries. A DO registry may manage metadata of objects from a certain repository. Another possibility is managing both metadata and actual digital object content stored by the registry, and a third scenario is managing metadata from multiple repositories. The DO Registry can be set for different types of metadata schemata and can be customized to provide different search, federation, handle registration, event management and other services.
The most popular application of this system is the use of Digital Object Identifiers (DOIs) to identify digital versions of written publications (e.g., scientific articles). Such identifiers, by means of an ID resolution, will lead not only to the digital object but also to its metadata. The important benefit of using DOIs are persistent citations (i.e., the location of the digital object may change over time but the identifiers will remain the same and its resolution will lead to the new location). The Linked Data movement has been pushing towards publishing and interlinking public data in standard formats, which enables the automated discovery, management and integration of structured resources online. The adopted technology is based on HTTP URIs and RDF. The resolution of an entity given its identifier boils down to three steps in that context: 1. discovering the IP address where the HTTP URI is supposed to be hosted (for example using the DNS) 2. contacting the corresponding server and negotiating the content (e.g., to serve a human-readable version of the RDF data if the client is a Web browser) 3. retrieve the structured description of the entity over HTTP.
This process is commonly called entity dereferencing since it is similar to general
URI dereferencing on the Web9.
2.4
In the context of the Okkam project, the Entity Name System (ENS) [
The existing data and entity registries are a critical asset for many Internet applications. However, all current architectures present limitations when we consider a situation with limited network connectivity or ad-hoc environments. If we imagine networks of computers with no connectivity to external Internet resources, it becomes clear that DNS-based entity registries typically do not work properly as only partial information will be cached in local and accessible DNS nodes. Moreover, data in the cache may not be up-to-date as DNS nodes may not frequently communicate with the rest of the DNS infrastructure. Similarly, clients may not be able to resolve DOI if the servers of the naming authority which issued such IDs is not reachable through the network. Entity registries like the Okkam ENS or LOD end-points are all based on centralized solutions which limit their reliability: in ad-hoc environments such central resolution points may or may not be reachable at a given point in time.
For the above mentioned reasons, we claim that a decentralized solution enriched with full replication of some seed content would be a better approach for an entity registry in ad-hoc environments. A decentralized system based, for example, on structured P2P networks can provide better connectivity thanks to cheap P2P communication and can tolerate the situation in which registry servers are not constantly available. Thus, in the following section, we envision solutions that consist of multiple distributed registry instances to optimize the availability of entity-registries in the ad-hoc environments. 4
We sketch out below three possible avenues that we have considered for downscaling entity registries in ad-hoc environments. Our solutions are based on IPV6, structured P2P and full replication technologies. For many, the first solution to our problem may be provided by IPV6 technologies. IPV6 guarantees directly-resolvable IP addresses, irrespective of the network topology (e.g., rending NAT irrelevant). Hence, local entities could be made available using local IP addresses only. One could even envision some address ranges reserved for such a use.
While practical and simple, we however think that solutions directly built on bare IPV6 technologies are limited for two reasons. First, we are missing the indirection layer offered by the DNS or similar registries, which is often essential when migrating or evolving entities (migrating a resource to another IPs would require setting up some redirection mechanism between the previous identifier and the new one). Second, from a networking perspective, this would boil down to mixing two separate layers, namely the networking and the application layers, which would certainly never be endorsed by the central networking companies and bodies.
Structured P2P Proxies
P2P technologies are another potential solution to our problem. Distributed
Hash-Tables (DHTs), such as Chord10 or our P-Grid system [
P2P technologies have been proposed in the past to enhance or supplant DNS
infrastructures11, most often to provide an alternative to ICANN or to support
P2P file exchange. Such efforts had limited success so far. We think that the
CoralCDN [
One could hence consider a DHT-based CDN as a starting point to solve
our problem, and enhance the infrastructure with a native entity storage system
(such as our recent dipLODocus system [
Entity Nucleus and Lazy Replication Even though supporting a full-fledged entity registry in ad-hoc settings is a necessity, there are many cases where some of the nodes might connect to centralized infrastructures intermittently. Thus, we believe that it is also essential to be able to cache authoritative or centralized information, and to be able to dynamically synchronize data with such infrastructures.
Depending on the context of the application, some core nucleus of the entity data can be identified (e.g., DBPedia entity data for LOD). In such a case and if the entity nucleus can be pre-installed on each node, then many of the entity operations can be resolved locally without resorting to any third-party infrastructure.
Networked search and updates, however, still require distributed mechanisms to be resolved. If such operations are deemed relatively infrequent, then semistructured P2P technologies like those described above can be applied: both distributed queries and updates can for instance be (lazily) propagated or broadcasted across the network at regular intervals. 10 http://pdos.csail.mit.edu/chord/ 11 see for instance http://blogs.computerworld.com/17444/p2p_dns_to_take_on_ icann_after_us_domain_seizures
Current entity registry solutions are often based on global hierarchies or centralized online directories. Such solutions are inapplicable to many contexts, including ad-hoc networks and environments that have limited access to a widearea connection. In this paper, we described some of the key problems related to using current entity registries in an ad-hoc context and suggested a few possible alternatives. Three potential solutions were specifically sketched, based on IPV6 technologies, scalable and structured P2P technologies, and (lazy) content replication. We now plan to implement, test and combine both current entity registry solutions and our new alternatives to determine in practice which architecture is most useful given a specific ad-hoc environment. As a start, we plan to focus on the OLPC XO context to provide a working solution and enable Open Data and Linked Entity applications for the billions of people who are currently cut-out of wide area networks. 6
This work was supported (in part) by the Swiss National Science Foundation under grant number PP00P2 128459.