W3C Delivery Context Workshop Sophia Antipolis, 4-5 March 2002 Jacco van Ossenbruggen and Lynda Hardman What we have learned -------------------- (besides that the French have a far more appropriate term for 'tourist attraction': 'distraction touristique', and that a small glass of chocolate-covered almonds costs 24 euros) Theoretically and academically, almost everything on the Web is related to device independence (DI). Practically, until now the only commercially interested party is the mobile industry. This resulted in the 2000 workshop in Bristol, http://www.w3.org/2002/02/DIWS/ http://www.w3.org/2000/10/DIAWorkshop/ that recommended the formation of the DI working group http://www.w3.org/2001/di/ (which we have now) and this workshop in Sophia that basically concluded that there are more DI issues than this working group can ever hope to solve. Device independence requires some notion of delivery context (DC) in order to be able to supply suitable information for the client. Delivery context can be broken up into familiar items: device profile; user profile; network profile; document profile etc.. The title of the workshop was "delivery context", but most of the motivation was from device independence. The difficulty with DI is that it requires an end-to-end solution, including at least: authoring support, server-side support, proxy/network-support, client support. So it only works if all these layers work and work together (weakest link principle). Current W3C specs that relate to DI include: - CC/PP for delivery context specification; - CSS media queries to define style rules that adapt to very high level device characteristics; - SMIL system test attributes; - HTTP content negotiation. Note that all specs use different vocabularies for overlapping functionality. This makes DI similar to i18n and WAI: a cross-W3C activity. In addition to W3C, IETF does DI related work on CONNEG (content negotiation) and OPES (Open Pluggable Edge Specification http://www.ietf-opes.org/), WAP forum has the UAProf (the wap phone User Agent Profile framework based on CC/PP). Technologies for doing the adaptation ------------------------------------- The position papers and the slides can be found at http://www.w3.org/2002/02/DIWS/final.html CSS (Bert Bos, W3C) using (CSS3 media) selectors. Michael Kraus, Munich, used browsing context which he added to the HTML file at the client to allow context adaptation. (He talks about nude documents, although naked would seem a more appropriate term for sensitive native English speakers.) Walter Dees, Philips Eindhoven, proposed a hierarchy of device capabilities with very high level capabilities at the top (e.g. audio interface vs graphical interface) to medium level (e.g. landscape vs portrait screen) to very low level (e.g. Nokia WAP Model X vs Nokia WAP model Y). Other people suggested similar ideas, e.g. to have several "layers" of device profiles and a way to map high level profiles (e.g. easy to use for the author) to very detailed (easy to use for the "last mile" adaptation done by e.g. the edge router of the network just before sending the presentation to the client device). If we are interested in the technologies and techniques, it might be useful to invite Michael Kraus and Walter Dees over for a visit. Walter is busy until about May 2002. Krishna Vedati's, Covigo, position paper wasn't very inspiring, but his talk was. http://www.w3.org/2002/02/DIWS/presentations/vedati/vedati.pdf Delivery Context description and use can be seen as a Semantic Web application ------------------------------------------------------------------------------ In order to adapt content to a particular device you need to know the characteristics of the device, and how to adapt the content for those characteristics. The first requires an ontology for the device description. You may have a device profile with defaults which can then be overridden by information "further down the line" (e.g. a proxy that converts html web content to wml for wap phones). This requires merging device descriptions and working out what the resultant description should be. This is in effect the same problem as combining different parts of ontologies. Additionally, if you have a description of the user profile and of the device, then you need to merge these. You also need to merge different sources (e.g. levels of detail, new model information) of device descriptions. All sound familiar? Andreas Schade from IBM said the group should talk to WebOnt about this as a use case. Same problems with deprecated terms in your ontology, and new attributes for new devices. Do you want to describe a basic core of properties, a la Tayeb Lemlouma, INRIA, then this is like trying to come up with a minimal ontology which everyone can first agree upon and then add their own extensions. Can this be done - or is it impossible to agree upon? (In any case, is it worth some of our time to incorporate some of the device characteristics in the Cuypers HTML entry form, rather than using screen size directly?) The same old arguments about RDF vs XML as a basis for the Semantic Web came up. Miscellaneous ------------- Adaptation and authoring discussion: http://www.w3.org/2002/02/DIWS/notes.htm#tuesdaymid This was a useful discussion. I (Lynda) have no idea whether the notes can be followed. The most useful part I found was a summary of techniques for device independence by Mark Butler (just before the end of the section): client vs (proxy or server) processing; can both linear and non-linear presentations (you can easily follow hyperlinks on a PC) be delivered from a single source; do you want to select from a given list of pre-made alternatives, or should you do automatic summarisation; also need to allow generation of different encodings, e.g. high-quality video to lower quality. It was entirely unclear what MPEG-21 is going to do. This could be because they won't be deciding until May. (http://mpeg.telecomitalialab.com/standards/mpeg-21/mpeg-21.htm) Link to the resulting MindMap: http://www.w3.org/2002/02/DIWS/mindmap.gif I (Lynda) think this is a linearisation of the map: http://www.w3.org/2002/02/DIWS/notes.htm#tuesdaypm ---***---