ICS 205 Software Evaluation
October 8, 1997
"Trellix: A whole new way with words"
Dan Bricklin's reputation was founded on the rock of VisiCalc, the
bestselling spreadsheet that finally made personal computing a
business fixture -- the original 'killer app'. He has been working on
pioneering general-purpose productivity applications ever since. His
other famed success was Demo, his toolkit for mocking up workable
prototypes of software and presentations.
His latest focus has been on reinventing the notion of document. The
result is Trellix, "The definitive productivity tool for the Web era.
It replaces the constrained, linear approach of traditional business
communications with Web-style documents -- complete with
color, graphics and linking. Readers can quickly scan summarries and
then drill down through the document -- or complementary Internet and
Intranet sites -- to obtain more detailed information. Trellix 1.0
documents can even be customized to guide individual readers to
This quote is from their web site, which promotes the free download .8
beta I report on herein. All I had to go on at this point is
Bricklin's reputation, a friend's recommendation, and some screenshots
of Trellix in action. One focus is a 'visual document map' across the
top of the window at all times.
I downloaded the application and a complete kit of sample documents.
Installation was just like any other modern Windows95 app:
InstallShield self-extracting .exe
Upon first launch, Trellix automatically offers a guided tour. It puts
up a full-screen presentation, with a menu across the top:
[Introduction | Map | Page Area | Links | Sequences | Reader Preview |
Samples]. The promised four-minute demo walks through that sequence as
the reader hits 'next'.
The first concept is that the main window is split vertically into a
document pane and a visual map of the document. The upper "map view"
has a page icon for each leaf: so I learn that a Trellix document is
like a self-contained 'mini-web'. Shuffling page icons in map view
also edits the structure of the document: related concepts can be
clustered and sorted and highlighted by 'topic color'.
The second pane is the 'page area' which contains one atomic chunk of
information. The tour hints at "masters and layouts" which automate
the look-and-feel of these pages. Graphics and other objects can be
embedded with the text, exactly like a modern word processor.
There are also up to four borders/margins to the page area Trellix can
fill in automatically with titles and navigational arrows.
It should be noted that their example document so far has very subtly
and effectively modeled the US Constitution as a Trellix document. The
map view orginally had thirty undifferentiated pages, which were
successively grouped, shaded, and labeled into preamble, articles, and
The third set of concepts, linking, is tougher. There are some
nonvisual mappings involved now. While editing, a control-click will
verify and follow a link. There are two kinds of link: basic and
title. A basic link starts from any selected region of text in the
current page view. At this point, the tour smartly offers to 'show me'
by simulating the steps involved. The trick is to highlight text and
then right-click the target page in the map view, and tick 'Link To'
off the object menu. The 'title link' it turns out, is similar, but
the text of the source label is taken from the title of the target
document -- and Trellix automatically tracks any changes to that
Fourth is "sequences". The explanation, simple enough, is that you can
draw a path through any set of pages in map view, and Trellix will
automatically add 'previous' and 'next' and 'up' and 'down' navigation
arrows to page views.
I found a hole in the demo: if you click in the 'application window'
graphic, a "This version of the demo is not fully interactive.
Please use the buttons at the bottom of the screen" alert flashes by.
Finally, there is a 'reader' mode that allows you to view the final
product. Editing menu options and toolbar items disappear, and the
page view is rendered in finished form.
At this point, it invites you to play with the sample documents.
I pulled up the Constitution example. All of a sudden I learn
something is fundamentally mutable which surprised me: the map view
can be split horizontally off to the side, too.
The map view works as advertised, though. I selected the first ten
amendments and created a new subsection of the map, the Bill of
Rights. I drag-selected the ten and experimentally right-clicked. I
didn't find any useful commands that way. [I'm trying a task that
wasn't illustrated in the demo: I'm trying to subdivide an existing
topic group] Next look is the toolbar. Most of the tools are
completely familiar document-management: open, save, delete. There are
two tools for 'view main map' vs. 'view overall map' -- what does that
mean? The other buttons are for links. Finally, over near the margin
(cut off by my 640x480 display mode: obviously the toolbar code had
not been tested to adapt to screen sizes), were "sequence" and "tour"
buttons. The visual iconography is no help yet, which is OK, since
toolbars are supposedly 'expert user' tools.
Having selected the ten amendments, I click "sequence", and something
rather unpredictable happens: a new outgoing yellow link *from the
entire Amendments topic*, and no change to the selected ten. At this
point, I discover there's no undo.
I click back on the miniature page maps, and I make a startling
discovery. I miss by a few pixels, and selectthe colored rectangle
that groups the Amendments together. With grow handles.
Yes: the graphical boxes and labels for chunks of a document are
purely graphic ornamentation on the map. I assumed they were
*structural* properties. Indeed, the paradigm on the left hand side of
the screen is a *drawing* tool. Now I can use my intution again. I
drag the ten away and prepare to draw a new box, and add a label,
"Bill of Rights". Unfortunately, there are no visible drawing tools.
I'm in 'select mode', so it's not clear how to drag out a colored
Turns out, it's on the "Insert" menu, not on any toolbar. I insert a
rectangle, which automatically comes out green, and a label, which
defaults to the helpful text: "Label: double-click to edit". Upon
double clicking, it disconcertingly changes font, but the operation
Now, I decide I'm not happy with green. How can I change the 'look' of
this document using those template they mentioned briefly? Inspecting
the menus, there's a "Format|Switch Page Layout" that seems promising.
Unfortunately, it pulls up a radio-list: "Choose a page layout for
this page: text or title page?" which isn't the effect I'm looking
There's another entry for "Format|Change Document Master" which I try
instead. Now comes choices of "Select a new master whiuch will be
applied to all pages in the document: Basic Green, Bold Red, Business
Financial, or Default". After some apprecialbe chunk of time, the
dialog goes away and I see the fonts have changed to a new look.
But not the color. I go back and right-click on the rectangle, and lo
and behold, there's the property sheet with a color tab, where I
should have thought to look in the first place. Again, in misguided
faith that there was strucutral meaning to the map view, and that the
system chose a new rectangle color automatically, I assumed the colors
were under 'master control' but were not.
I tried some other experiments: I still haven't figured out how to
create a free-standing sequence of, say, amendments 13-20. I haven't
felt compelled to test any of the text-editing features, though, since
it's hard to imagine them operating in any other way but the expected.
I think a bit about my workflow and hypothesize I'd like to be able to
clip 'post-it' notes into a Trellix notebook throughout my day. This
requires much tighter integration with mail, my web browser, and
calendar. The web integration should be there, but I haven't tested it
yet in 90 minutes of notetaking.
But the value would only be if Trellix could use its UI to help me
*organize* the information. The map view is just less powerful than my
fantasies imagined it to be. I can't seem to do queries, grouping by
similarity, automatic maintenance of a clean, pretty map, and so on.
I'd love to keep a chronological timeline with color topic coding,
etc. The map itself is as important to my web development habits as
the html editor for the leaf pages.
I also haven't discovered any reader-discrimination features yet. I
can see the value of creating different tours through a programming
textbook for say, C and Fortran programmers.
The last test is exporting my modified constitution back out as HTML.
Saving the .tlx document first takes more time than I thought, making
me suspect a crash. The window freezes, graphics tear, but eventually
it comes back to life and I "Save As HTML". After an even longer wait
-- apparently file operations are in a blocking thread that prevents
redraw -- it's done. Closing Trellix, I am a little unhappy to see
176 html files littered across my desktop. I *assumed* it woiuld
spit out the results in a separate folder.
Final analysis: I spent my undergrad years writing a hypermedia
textbook authoring environment for NeXTSTEP that converted to HTML,
LaTeX, text, RTF, etc, and I aimed at building the kind of application
Trellix is today, four years later. I am floored.
1. List some application functions. What's most important to you? What
won't you use? Will most users use the same set or different sets?
For assembling quick miniwebs, some text, and a pasted chart or
graphic, I think it has all the obvious functions. The map is more
important to me, but I forsee high overlap between most users. It's a
single-purpose general-purpose tool :-)
2. Describe the UI in general. Preferred input channels? Are the
displays simple or complex? Is color, sound, or animation used? Is
there help or other documentation?
The UI aims to be as familiar as possible to Windows95 office
productivity application users. There's a toolbar for the standard
document features across the top, a main editing pane below that's
ready for typing and drag-and-drop.
As a navigational tool, it uses color througout and relies on the
mouse, though it has extensive keyboard equivalents from Win32s.
I used the web site. No printed documents or other magazine reviews
were consulted (though I read the reviews later). It had online help,
which I did NOT use in this experience.
The sample files constitute a sort of documentation, too. It's often
overlooked in the sw dev process how training is part of the UI
problem, and how tacit artifacts like examples in the documentation
shape user's expectation. The Constitution is so structured I was
misled by map view.
3. Narrative of trying out the application: see above. Summarize one
nice thing about the interface; and how it can be improved
I really like the linking model. It's as good as intra-application
linking can get. Interapplication linking is still a morass on most
operating environments, though. And it's simply not possible to link
*to* a selected range of text (though that's completely possible in
HTML). The only better solution is to train users to visualize a
separate 'link clipboard', which then offers full flexibility, as
shown in Tim Berners-Lee's original NeXTSTEP NeXUS Web browser/editor.
Toolbars should be modal. When I was in the map pane, a huge useless
strip of text-editing tools was in the second row, when I think there
should have been a map-specific toolbar that was swapped in. Instead,
some (only some!) map tools were lumped into the upper document toolbar.
Undo is not yet supported. Yes, it's completely understandable for a
Sneak Peek, but it is REALLY critical for an application as central as this.