The first workshop on Downscaling the Semantic Web took place on May 28 during the Extended Semantic Web Conference in Heraklion.Following the initial public recognition the idea of DownScaling the Semantic Web received and a initial set of related initiatives (Voices, SemanticXO, …), the goal of this workshop were to: bootstrap a community around this research topic, more precisely identify the challenges to overcome and assess the current situation. We accepted 7 contributions and had a lot of interesting discussions during this day.
What do we mean by DownScaling ?
Currently, the main stream approach to the design of a data platform is a centralised hosting read/write solution and lightweights clients. Example of these include Wikipedia, Facebook, DBpedia, … All the issues related to data integration are solved centraly and data access is ensured on the basis of a reliable infrastructure. DownScaling means providing a alternative, non-centralised based, approach to this. For example, the traditional mail system deployment with different servers and applications is a DownScaled version of mailing as seen by centralised mail portals such as Google Mail. Another example is Bittorrent has a DownScaled alternative to data sharing platforms such as RapidShare. DownScaling does not however implies breaking totally with the centralised model, DownScaled data centric platforms can rely on centralised solutions for part of their needs – although this may come with some constraints on the applicability of the DownScaled solution.
There could be several reasons for which one would like to part from a centralised model. For instance:
- Accessibility: centralised solutions rely on the availability of a strong infrastructure for both hosting the service and delivering the content. Citizens having no access to such infrastructure are thus deprived from making use of them. DownScaling is a way towards dealing with the lack of infrastructure for accessing centralised services by providing alternatives making use of a decentralised model.
- Privacy: DownScaling the centralised systems into smaller units, eventually up to the user level, allows to achieve a mapping of 1-to-1 between the user who produces data and the device that shares it. This potentially increases the level of control and ownership that user will have on his data.
- Cultural habits: central access point may not be compatible with the usage and habits of some social interactions. For instance, community radios shows a different interaction pattern than national broadcast and are closer to the population they broadcast for. Information systems can be DownScaled to the community level to mimic this behaviour. Another example is that of communities used to a particular system and communication codes. As it is hard for these communities to give on their habits to move onto a centralised system with homogenized rules interoperability solutions have to be designed with a DownScaled view in mind.
What are the problems to tackle?
These are the problems we identify to be in relation to DownScaling in general or to some more specific contexts of its application, like that of developping countries:
- Constrained hardware: DownScaling the software solutions eventually goes with DownScaling the computing power as well. Either because the higher performance hardware is not available or because it is not needed – splitting the load reduces the need for resources. There is also the added external constraints such a limited Internet connectivity and lack of electricity. DownScaled solutions have to be able to operate under constrained conditions.
- Data query/filtering/availability/trust: Switching from a central server to a swarm of them rhymes with an increase in problems related to querying, filtering and trusting the data. Let alone that of having access to it. Several servers could potentially be providers of some information sought by another device. If these servers are low scaled and operating with limited online connectivity, the data they provide will have limited availability. The target here for DownScaled solutions is to be able to maintain a good quality of service despite this data related issues.
- Cultural issues and costs: The access to technology and its usage, DownScaled or not, is very much bound to the cultural backgroung and skills of the target users. This is even more so when the technology is brought closer to the citizens. When DownScaling a global system down to a community level, this system has to get closer to this community. Different community may most likely speak different languages, expect different from of interactions and have varied litteracy levels. DownScaled systems have to be able to cope with contextual peculiarities that appear at the scale they operate in, while being able to maintain a global cohesion within the system at large.
Where are we now?
The participants of the workshop came from different research and industrial entities working on different aspects of the problems we collaboratively refined. Some researchers are focused on the interaction between emergency response teams from different countries, thereby looking at dealing with different local habits within DownScaled information systems. Some other are focused on voice-based access to information to cope with illiteracy issues and de-centralised model for information reporting based on the community radios. There are also other teams working on more technical aspects directly related to data access: the connectivity with volatile data resources, the exposition of data contained in mobile phones and query execution under constrained conditions. Finally, we discussed some general interest in the ethical issues around improving access to data in developping countries.
Besides the participants who came to discuss their papers, we also enjoyed the presence and the support of other conference attendees which however they did not contribute to the proceedings where interested in thinking along with us 🙂