During my vacation back east in May and June, I spent two days learning about computing in two other organizations. I also discussed computing with colleagues at Cornell and Princeton Universities.
The theme of this year's Open House was "Information Abundant Interfaces." During the morning and early afternoon, current HCIL research on visual interfaces for presenting and using effectively massive amounts of information were given. Demonstrations completed the day.
Interfaces need to become comprehensible, predictable, and controllable. Topics that particularly caught my attention included ways to improve information seeking and browsing, refinement of information visualization, and specification and evaluation of interfaces. Emerging guidelines include:
This approach was used by some undergrads who brought up in a month the "Dynamic Home Finder" based on the Washington, D.C., area real estate listings. The display starts with a map of all houses for sale. As one enters restrictions, e.g., location, size, limit on lenth of commute, and cost, only those houses meeting the selection criteria continue to display. When one sees a house one's curious about, one can request its specifics. If one decides there's nothing left that's interesting, one may loosen any of the selection criteria and update the map.
The students encountered a lot of resistance to this idea. To show its power, they held a "data entering party," put in 1100 houses, and made their point. Now the application is catching on among realtors. This is one instance of the "zoomable scattergram," which seems widely applicable.
Zooming and filtering have also been found useful for dynamic exploration of data to investigate trends and identify outliers. Showing multiple levels of view at once may help maintain context. Using color and thickness to indicate characteristics also seems promising. Dense screens are powerful, but difficult to design.
Tools to evaluate interfaces like HCIL's Questionnaire for User Interaction Satisfaction (QUIS) and Automated Consistency Analysis are coming of age. Ethnographic methods and anomaly classification seem to have potential. Templates can help generate consistent interfaces.
These text findings and the power of dense screens probably shed light on the controversy surrounding the layout of the SLAC Home Page.
Another useful session was "Spinning Your Web: WWW Interface Design Issues" by Vince Boisselle. Before designing the pages on the Broadcast Pioneers Library for his server, he reviewed the hypertext literature, finding similarities and differences between self-contained hypertext and the global connectivity of the Web. Quoting from Shneiderman, Boisselle noted:
The intrigue of hypertext is the replacement of the simple linearity of traditional written text with the opportunity for jumping to a variety of successor articles or cards...Differences between traditional hypertext and the Web that Boisselle emphasized include:The dual dangers are that hypertext may be inappropriate for some projects and that the design of the hypertext may be poor (for example, too many links, confusing structure)...
Some of the sessions are summarized on a video tape given to the attendees.These include:
Wish we had one! It's an interactive learning environment, set up so each pair of students may share one PC. In addition to computers, the Lab uses AV and teleconferencing equipment. Among things demo'd was the anonymous discussion by students via "chatting" software of topics proposed by the instructor. (IBM, etc., studies have shown comments are weighted differently when the visual characteristics of the speakers are unknown.) The University has found it pays to have a technical support person available to the instructors at all times.
My "paired" student, Adlai Waksman, pointed me at his colleague's page with the "Darwin" Fish, which he got from a Peninsula store. (It's a small Web world!)
This turned into a group discussion of Web design and maintenance issues. The consensus among the attendees who'd seen it was that we liked the Yale Style Guide best so far.
Vince Boisselle said he had found the HotMetal HTML editor more trouble than it was worth unless he's creating a document from scratch.
People thought menu bars especially require the choice of very precise terms. Collaboration with a graphic designer was recommended. People thought the Princeton University Library menu bar (near the bottom) set a good example.
Several librarians spoke to the need to re-establish online the intellectual controls for accessing material that were previously done physically. Part of understanding material is to apprehend its context. Previously, one learned that by physically visiting the library. How do we re-establish context online?
People also pointed out we need to think about "access for what?" Sometimes it may be enough to publicize that a particular item is available at a site, rather than having to put the entire document online (though this is certainly not what we've found with physics e-prints...)
REXX is the most popular language at the Library of Congress for user-written CGI-bin scripts. (User knowledge of REXX came from OS/2.) Professional programmers use a C programming language instead. Perl is not used much at all.
Some felt not knowing who the users are of one's WWW server is a "security flaw;" but as had been brought out in an earlier session, there is a well-known document, the "Belmont Guidelines on Research Involving Human Subjects." This requires the informed consent of all subjects unless their identity cannot be determined from the data. If the individual userIDs are somehow masked, transaction log analysis of individual users could be beneficial in understanding the strengths and weaknesses of one's Web site.
NIST is working on guidelines for federal government research involving information technology and user interactions. The Direct Marketing Association (DMA) already has fairly stringent guidelines.
A person from MITRE has found it productive to mirror his entire page hierarchy in test and production versions.
Several emphasized the importance of setting up one's information hierarchy very carefully. In addition to helping users find their ways around, a well-thought-out architecture is important for maintainability.
This was at the new Neutral Buoyancy Research Facility of the Space Systems Laboratory. It was much easier to manipulate the satellite in the neutral buoyancy tank telerobotically than to manipulate the computer simulation. While waiting, I got to try out some "crystal eyes," glasses that showed a live TV view of the wearer and environs in stereo. Pretty realistic!
I wish I could have seen more. Getting from place to place was a problem. I have a number of short reports published recently by HCIL, too, in case anyone's interested in reading them, along with other bibliographical material.
While there, I demo'd the SLAC Web to their postmaster, Jim Hubbs; and he showed me both the APL internal and new external WWW page sets. APL has put their internal pages on one server and their ones for public consumption on another. At least some realize this does not meet the needs of collaborators who are not on-site, but it's working more or less so far. Communications' Greg Hollingsworth, to whom Ed Hart introduced me, led in prototyping WWW pages to prove their potential usefulness. TechPubs' Lucinda Halbrook (who was unfortunately not there) has been working to create polished, public ones.
Another colleague, Tony Waters, gave me a short demo of the Work Breakdown Structure part of a Funding, Planning, and Budgeting client/server GUI application being developed by APL with Price Waterhouse. It's for the central budget office and traces the money flavors and flow. They're using PowerBuilder on MS/Windows (just recently available on Macs, too) to develop an application that communicates with an Oracle data base. Tony is quite happy with PowerBuilder and finds Oracle to have more functionality and be more intuitive to use than Nomad, the 4GL database query language they previously used to access Rdb. To layout their GUI screens, they are using an in-house standards document that applies to both MS/Windows and Macintosh environments. (Whether someone has a PC or a Mac depends on the particular APL group's culture. Both kinds are common at the Lab.)
Tony is also in charge of a committee to obtain a Document Information Retrieval and Indexing System (DIRIS) to hold, index, retrieve, and present compound documents (e.g., a Word document with imbedded spreadsheet). A DIRIS Web server may be used to generate and serve HTML-format output on the fly. I obtained some information about APL's requirements as well as a fancy, home-brew, KWOC application they have been using that includes an extensive stop word list, synonym processing, and "go" list (where extra keywords are "attached" to the titles for indexing). I expect the complete requirements, name of DIRIS product selected, costs, etc., soon.
cron, disaster recovery planning, and a FAQ.
TO_CHAR('HH:MM:SS')
does not mean hours:minutes:seconds
but rather hours:months:seconds!
This caused a colleague quite a bit of time debugging.
To get hours:minutes:seconds, code:
TO_CHAR('HH:MI:SS').
I have copies of the two ADSM manuals in my office.
Winters