The Trellix 1.0 development project
An overview of the history of the project that helped us come up with the ideas presented here.
Last modified: 9/28/98
The development of the Trellix 1.0 project provided a lot of input to the philosophy followed here.
Trellix 1.0 is a desktop application software product for creating web documents. It is designed specifically to create the type of documents described here (hence our interest in the subject...). See www.trellix.com for more information.
Trellix 1.0 was used to create the original version of this web site. Trellix 2.0, a later release, was used to create the current web site.
Here is an overview of some of the early part of that development:
First, we decided that we wanted to create a tool for writing linked, web documents. Then we got samples of many documents used in business. We chose one representative sample, a consulting company's evaluation of a product, and got the original word processed source on which to experiment.
We decided to use Microsoft Windows Help as our linked document environment. There were tools available, and the formatting was compatible with our original (this was early in the days of HTML).
We tried converting it to a linked document by breaking it into sections. The resulting document was very hard to read.
We then tried making the document more hierarchical, with overview pages with links to the detail. A little better, but you had to follow a link to learn what it pointed to. We tried finding words we could use from the original document for titles and summaries at the link site.
It became clear that we needed to write the summaries and titles explicitly for linking. We couldn't mechanically use the text of the original. Worse yet, we had to really understand the document's contents to write them well... This is where we learned that you really do need to write explicitly for a linked output.
The document still had a problem: We always felt lost when reading, not knowing where we were in the document nor its structure. We tried using nested titles for the page (a typographic navigation-bar solution), but, unless you knew the document well, that didn't give you a feel for where you were. The person who wrote them felt it did, but not those of us who didn't write them. Also, you couldn't jump around globally like you could in a paper document.
We then came up with the visual map of the document, with very small icons for each page. Landmarks (other text and shapes) were added to make the map understandable. Finally, we found that indicating which page you were currently viewing ("you are here") and making the map a clickable imagemap worked great. (Unfortunately, this required almost 50 different images in the prototype...).
Finally, we had a document that worked.
We also experimented with tours, the wording of links, and other techniques.
Bringing the document back to the original author, we were surprised (and pleased) when he said that he felt it was more readable than the original paper document. (He actually said, "They'll finally read my document.") We tried it on others, and they liked the concept. (So much so, that it was the basis for the first round of funding for Trellix Corporation.)
We continued to prototype based on real documents. From this we honed various layout techniques, including the mixture of scrolling and non-scrolling areas.
In addition to this experimentation, we brought in some consultants to help in our design. Chris Daly, one of those consultants, is a journalist, and helped us develop the newspaper-centered ideas presented here. (Chris also teaches journalism at Boston University -- he has a web site at www.bu.edu/cdaly.) The document he wrote as part of his work for us is on the Web at www.bu.edu/cdaly/hyper.html (it's a linear document since it was originally meant for print).
Once we were into prototyping and then development of final code, we set up a usability lab and brought through nearly 100 people to help us polish the product before shipping 1.0. Further testing happened during the development of Trellix 2.0.
All of this work, writing summaries, formatting pages, looking at how people read, watching usability tapes, etc., and lots of introspection, helped us come up with some guidelines.