Getting Traction

Traction's enterprise Weblog system gets a grip on corporate KM

THERE IS STILL NO sure-fire recipe for KM (knowledge management) success, but the ingredients must include the staples of the knowledge worker: e-mail, the Web, and Microsoft Office. With Traction Software's KM solution, content flowing through all these channels is easily captured by the Java-based Traction Server, which can be best described as an enterprise Weblog system. Documents posted to the server are stored as XML, tagged (in a Web interface) with system-and user-defined terms, made available for full-text and structured search, and served back out as team workspaces, enterprise information portals, or both.

Traction says you'll be up and running in 15 minutes, and that was true of our Windows XP installation. We spent another 15 minutes configuring the server for SSL (Secure Sockets Layer). To do that, we downloaded Sun's JCE (Java Cryptography Engine) and JSSE (Java Secure Socket Extension) JAR (Java Archive) files, dropped them into the server's directory, and ran the canned keytool command provided by Traction to create a self-signed certificate.

New to Version 2.8 is the automatic enabling of a visitor (anonymous) account. We exploited it by creating two Weblogs, or "projects": a public one for visitors and a private one called watercooler, where a team could assemble, discuss, and refine content to be made public. We used Traction's Web-based administration to define users and permissions. There's also an LDAP provider interface for synching enterprise directories, which we did not test.

Logged-in users add, edit, classify, and erase articles using a Web form that accepts HTML from any source, including e-mail attachments. The system provides a set of default labels -- including headline, news, notice, to-do, done, bulletin, contact -- and invites authors to apply these to posted articles. Labels define rules for displaying articles. For example, headlines come first on the portal's news page, followed by news items, then notices. All of these obey the current time slice, and therefore are visible by day, week, month, or quarter. Bulletins, however, are time-independent and always appear on the news page.

After they are posted, articles or paragraphs within them can be assigned more labels. For example, a paragraph of contact information can be tagged with the contact label. A search for the label returns all documents containing paragraphs so tagged, and highlights the paragraphs. Users can easily extend the set of system-defined labels. While testing, we created a "howto" label for Traction lore as we acquired it, and a "wssecurity" label for Web pages and e-mails related to a research project on Web services security. Even without such labels, these documents can be found using Traction's built-in, full-text search engine. But the system is designed to encourage and reward metadata tagging.

Search is menu-driven for novices. But for experts there's a powerful and terse syntax. For example, /:howto finds all how-to articles in the current time slice, and /::news (saml and ws-security) w finds all news items for the current week matching SAML (Security Assertion Markup Language) and WS-Security. A related syntax makes it easy to link to these dynamic result sets. To give users easy access to our series of how-to items, for example, we posted a bulletin containing the text [[link :howto 'howto' items]] that Traction rewrote as a link, labeled "howto items," that invokes the query and returns the set of items within the current time slice.

Mastery of such syntax is, admittedly, a power-user's game. Novices (and rusty experts) require WYSIWYG support, which Traction's Web interface delivers for search but not for content authoring. A companion Windows tool, the Traction Instant Publisher, embeds the Microsoft DHTML edit control that it uses to offer basic WYSIWYG editing of new entries it posts to the server by way of a remote API. It's an HTML editor that supports simple styles and hyperlinking (but not tables), has access to the same labels available in the Web interface, and can search the server as well as post to it.

But the Instant Publisher can't retrieve, edit, or reclassify items that have been posted, and it doesn't simplify use of Traction's link or search syntax. A friendlier interface for novices, who will otherwise struggle to master the deep and powerful Traction engine, should be a top priority in future versions.

The Instant Publisher is, nevertheless, very effective in its primary mission, which is to push raw material onto the server. Running from the Windows tray, it senses the URL in your browser, the e-mail message selected in Outlook, or the document open in Word, and allows you to post any of these with one click. It also includes a handy screen-capture tool.

E-mail is another vital source of raw material. We set up a POP account and told Traction to poll it and move messages into our private project, labeling them as news items. Conversely, we told Traction to send e-mail digests summarizing project activity to team members. This two-way e-mail flow is crucial for adoption in companies that rely heavily on e-mail and on the RIM BlackBerry in particular.

There's much more to Traction than we can say here. The company has deep roots in hypertext, and its engine weaves a dense fabric. Every action is recorded and cross-referenced. Every document's inbound and outbound links can be displayed. Items can be collected and reclassified in batches. The interface is highly skinnable in ways that affect not just appearance, but the level of exposed detail. There's no lock-in to the Traction repository; all data can be easily exported as XML. To create an ad-hoc RSS (Rich Site Summary) feed, you add &skin=rss10 to any view-generating URL; to retrieve search results as XML, you add &theme=xml to any search URL.

Traction can be regarded as an ultrasophisticated PIM (personal information manager). That's just the quality you need in a KM product to get people to actually use it.