Downtown Banff

New material added May 24, 2007.

Trip Report - W4A/WWW

May 7-12, Banff, Alberta, Canada

"Magic is stuff you do not understand, yet." - Sir Tim Berners-Lee

Two Conferences In One

Attending these two conferences meant six solid days of multi-track presentations, but it was fascinating. This was W4A's first year as an actual conference (it has been an interest group associated with the WWW annual conferences for years). Primary themes in both conferences included the following:

The following are notes on talks that seemed most relevant to the UW.

Accessibility

Accessible Rich Internet Applications (ARIA) - Becky Gibson kicked off W4A with an overview of work on enabling the accessibility of Web 2.0 sites. Becky described an impressive level of cooperative working between Google, Microsoft, IBM, and W3C to define and apply the Accessible Rich Internet Applications (ARIA) methods to interactive Web interfaces, including the development of the Dojo JavaScript Toolkit.

Enhancing for Accessibility - UW's Jeffrey Bigham and Richard Ladner described AccessMonkey which, like GreaseMonkey, creates Javascript enhancements of Web pages. AccessMonkey could be used to create more accessible navigation on a page, for example. The concept is to empower disabled users to add useful functionality to pages they are interested in. Communities of users with common needs can share the functions. Site developers can keep an eye on what such communities are developing for ideas on how to enhance their sites. This approach had me musing on some interesting questions:

Supporting Accessibility - Microsoft's Shelly and Young sorted out levels of complexity in Web pages that affect accessibility and described specific coding approaches for moderate complexity DHTML. Their presentation broke Web pages into four levels, with suggestions on how to support accessibility at each level:

Accessible Mathematics - Bernareggi and Archambault described progress in making mathematics accessible to the blind. Their primary point is that MathML is an excellent approach to making mathematics accessible. One challenge is that many different kinds of braille exist, particularly in how math is represented, but MathML can be converted to most of the common forms of braille.

The Importance of Headings - Watanabe presented a classic usability research paper on the value of a good structure of headings in a Web page, demonstrating that headings can sharply reduce task completion time for blind as well as sighted users.

Accessibility Into The Future - Kelly et al suggested that the future of Web development, with its growing complexity, will require that achieving accessibility be less a rules and criteria process and more of a cooperative process among stakeholders. Kelly characterized the shift as a move from "cathedral" (idealized, centralized, and heirachical) to "bazaar" (decentralized, pragmatic, democratic). "The focus will be on the journey, not the destination," said Kelly.

Interesting Tools - Several interesting tools were mentioned in the course of the presentations:

The World Wide Web

Generally I went to sessions relating to small screen and mobile Web design, security, usability, and accessibility, with some side-trips to Semantic Web stuff. Here are some of the more interesting talks:

Academic Ratings - Working to go beyond the Thomson Scientific ISI Impact Factor method for evaluation scholarly impact, Bollen, Rodriguez, and Van de Sompel of Los Alamos National Labs are building a semantic network integrating bibliographic, citation, and usage data of scholarly publications to develop new metrics.

Defeating Injection Attacks - Jim, Swamy, and Hicks described a browser policy system for blocking injection attacks, either from forms or in the Web pages themselves. The fun(?) part of the talk was the long list of examples they offered of methods for doing injection attacks in today's Web of wikis, blogs, and social environments, concluding that reliably detecting injection attacks is very difficult. The cleverest example was a Web page with embedded code that, by itself, was a nonsense string of characters. Errors in the HTML in the page would make browsers process the page in quirks mode, which would "clean up" the HTML before running it, thus converting the string into functional scripting which then attacks the computer.

Early Detection of Phish - Several speakers described methods for recognizing phishing messages. Given the short life of a typical phishing site, black lists are not too useful for spotting new stuff. Zhang, Hong, and Cranor described a method that scans suspicious messages with the Term Frequency - Inverse Document Frequency (TF-IDF) algorithm to identify important words and then runs the words in Google. If the orginating domain and all URLs in the message rank high in the search, the message is probably legitimate. Phishing sites are usually not around long enough to be crawled and thus will always get low ranks or not show. Another technique described by Carnegie Mellon's Fette, Sadeh, and Tomasic is to check the age of domains in the message against its date. If any domain related to the the message was created about the time the message was sent, the message is probably phish. The Carnegie Mellon Usable Privacy and Security (CUPS) Lab site has papers on other methods.

Mashing Up Web 2.0 and the Semantic Web - University of Karlsruhe's Ankolekar, Kroetzsch, Tran, and Vrandecic presented a position paper asserting that "future Web applications will retain the Web 2.0 focus on community and usability, while drawing on Semantic Web infrastructure to facilitate mash-up information sharing." The approach could give Web 2.0 access to information of far greater depth and breadth than anything currently offered in today's cute environments.

Going From Mash-up to Hook-up - Many discussions took place on how to sort out the cacophony of rich media, dynamic page content methods. "Using AJAX is like doing Web page design in assembly language," commented one speaker. IBM's Rafah Hosn described the XML Application Components(XAC), including the Rich Web Application Backplane, which are intended to provide data-centric, collaborative, simple, and remixable methods. X-port.net's Mark Birbeck talked about his work with xH, a standards based Web Application programming language that integrates XHTML, RDFa, Xforms, XML, XBL, and SMIL. More information about xH is on the FormsPlayer Web site.

Mobile Web

Mobile Web - The Web is rapidly expanding into mobile devices, which were variously defined as "devices with really small screens," "devices with less memory and CPU than prefered," and other phrases. W3C has a Mobile Web Initiative and has recently published Mobile Web Best Practices. Nokia's Art Barstow talked about technical directions of mobile technology, predicting increased AJAX and widget use and emphasizing the importance of separation of content and presentation, valid structured markup, and semantic elements for any page intended to be viewed by a mobile. Page designs with AJAX and other scripting and widgets are coming, keeping in mind that the processing may take place partly on the mobile company's server, with simplified content being sent to the mobile device. Barstow said he felt proliferation of mobile-specific methods was bad practice - developers should stick to standard "full HTML and CSS".

MobileOK - Vodafone's Don Appelquist talking about MobileOK, a set of Mobile Web best practices worked out by the major mobile service providers. He gave me a handful of little MobileOK quick reference cards (if you would like one, let me know).

Goals and Bad Practices - Nokia's Art Barstow talked about goals and bad practices. In particular he advised against mobile-specific technologies. He suggested instead using "full xhtml/css", but doing so cognizant of the limitations of mobiles.

Best On Any Device - Opera's Charles McCathieNevile talked about work Opera did on its mobile browser Operamini (http://www.operamini.com/) and how to provide the best Internet experience on any device.

Mobile Domain - .Mobi's James Pearce described the work to create a new top level Internet domain named .mobi. The site dev.mobi has extensive resources for people developing Web sites for mobiles. Pearce was a walk-on, so I do not believe his talk is available online, but the http://dev.mobi/ site has tons of stuff and tools.

Other Mobile Comments:

Musings on the Balcony

The conference site has a wonderful balcony overlooking the Bow River valley where people tended to gather at the end of the day. Many wandering conversations occurred there. Several informal chats kicked around the topic of what all this could mean to higher education:

Social Networking in Academia - Several presenters have been trying to probe the dynamics of social networks and what educational role they could play.

This presents some puzzles for the use of social networks in higher education. While it is common for a small portion of a group to be more active, by definition in education every individual needs to be productive in a way that can assessed in terms of gained knowledge and skill.

Adding a generic social network to a program's Web services may just result in a lot of lurkers and a few mavens with little connection to educational work, particularly with the competition from other social sites.

One idea discussed on the balcony was creating a "peer garden", sort of like a theme tavern:

The goal is to mimic the professional relationship between work and participating in a peer community.

Connection to Information - Semantic Web activities such as the MESUR project, which integrates bibilographic, citation, and usage data from scholarly publications to develop metrics of scholarly impact, all within the OWL scholarly ontology, hint at some of the possibilities coming our way. Huge, active, structured, documented datasets in frameworks well equipped with tools for exploration and analysis could be connected to Web 2.0 UI methods to deliver a rich information environment to researchers and students alike.