Rohit Khare (rohit@fdr.ICS.uci.edu)
Fri, 21 Aug 1998 03:12:28 -0700

So, in one of Dave Kristol's emails to HTTP-WG, he pointed out that Jeff
Hostetler, late of Spyglass, now works for abiSource, Inc. What's that?


...obviously enough. Eric Sink and presumably others from the (flushed)
Spyglass team (itself based on what would now be called "grey" source from
NCSA -- Telnet, Dicer, Mosaic) have launched a company specifically aimed at
desktop applications.

The first out the gate is abiWord, to be announced in a few hours at the
O'Reilly/Open Source confab today (gee, I wonder if there ought to be an Open
Source Journal :-)

It's only a developer alpha release, naturally enough -- you can't announce a
finished Open Source product, by defintion.

Aimed at Linux and Win32. Mac loses again... They're writing fresh
cross-platform libraries, but only those two as strategic choices. I don't
fault them for that at all.

[Perhaps Apple should take on the mantle of aligning Mac open source hackers
and putting prize machines and, more importantly, Apple ad campaign fame on
the table]

A nagging indicator: the initial press release talks about the *functionality*
of a word processor-to-be for a bare paragraph out of pages about how cool
this new business model is. Folks, I need a word processor, not a philosophy
-- at least, that's the endgame: measuring by features alone, not grading on a
curve because a product's "politically correct."

A second nagging indicator? There are two features demonstrated in their
screenshots -- illustrating the needs of the "super-nerdy" perfectly. 1)
Multicolumn editable text (nice enough) and 2) a circular column (freaky deaky
developer backflips).

A third nagging indicator? The feature roadmap is a wishlist, naturally
enough. It's not the kind of rigorous Project Manager warplan you'd expect
from Steve McConnell's MS Press books (Code Complete, &c). The import/export
is one of the most ambitious, and some of them are going to be very lossy,
indeed: many of the features that make those documents work are not in the
target abiWord set.

And yet -- I'm pretty pumped about this effort. It's a Good Cause (TM), it's
run by Smart People (TM), and it's got a scalable goal set -- not to be
everything by version 1. This is the Next Step in the open source movement:
deploying paid, professional product managers to own features and progress, to
coordinate the feedback from the world to a coherent end.

That said, one can (and must) ask of any project: Does the world need another
word processor? (translation: "is MS Word a natural monopoly?") Yes, Open
Source OSes need desktop apps. No, Microsoft is NOT going to port theirs. So
we have a quandary: as meager as the market is, and as much as your "best"
users of all those features -- people who live and die by word processing --
have moved to the closed platform already, how can abiWord mature beyond a
hacker's quickie file viewer?

See, part of the Open Source magic is aligning user-developers' interests with
one's own. For systems-hackers, it yields programmable editors like emacs. For
the scientific-hackers, it aligned around TeX. For the latest generation of
Web-hackers, it ought to align around pure XML+XSL+CSS+PNG. The first two
development paths have produced hypertophied specimens: awesomely powerful and
aggresive entrants that destroy any competitors that come near: emacs owns
coding (with vi) and TeX owns scientific documents (with FrameMaker).

No one owns the "web document" -- in large part because it hasn't standardized
yet. There appears to be great promise, though, for building a fresh
interpretation of the office suite using the "work web" as the basic user
metaphor and development architecture (see also Trellix).

abiSource doesn't seem to be doing this. Yet.

The beauty of the bazzar is that if the hackers want to make the ultimate Web
tool, they can -- by dint of effort alone, they can control the architecture
of the product and direct it towards those standards.

If abiSource hasn't specified a roadmap for its internal implementation
architecture (so far it's just the features), that's *OK* for an Open Source
company. They don't own it; they share it. But once it ramps up, I'm hoping
the abiSource PMs ride herd on it and hold to that architecture.

Because, after all, the bazaar in every town square operates in the *shadow*
of the cathedral. (*)

Rohit Khare

(*) translation: *someone's* gotta design the thing...

The FAQ is very interesting reading. Excerpts:

Why are you doing this? If AbiWord is free, how will you make money?

It is our intention that AbiWord will always be Open Source. We will be
selling technical support, packaged CDROMs, manuals, companion products, and
other resources.

Why does AbiWord use XML as its native file format?

Because we like XML. Seriously, it seemed silly to invent yet
another file format, when
XML provides a very nice syntactic structure for our use. There
are a variety of other
options that didn't make as much sense for our purposes.

Why doesn't AbiWord use HTML, XSL, CSS, DSSSL, etc?

HTML is a web page language. It lacks the features necessary to
represent many word
processing constructs. Shoehorning those features into HTML
would be possible, but not

CSS was used as the basis for our design. However, CSS2 has
weak support for
pagination, and its box model breaks down somewhat when
designing algorithms for
non-rectangular text flow.

At the time of this writing, XSL was a moving target. We may
end up moving in that
direction over time.

DSSSL is far too complicated for our needs. As for style sheets
in general, AbiWord is a
word processor, not a structured XML editor. Separation of
semantic markup from
formatting instructions was never a goal.

Under Linux, why are you using GTK+ instead of Qt?

Politics. Most of the truly nerdy open source people prefer
GTK+, since Qt is not quite
free enough for the deepest dogma. We want the enthusiasm of
those super-geeks
right from the beginning, so we are making the choice that will
make them happy.
Truth be told, we would like to support both. If you would like
to help with a Qt port,
please let us know.


Rohit Khare -- UC Irvine -- 4K Associates -- +1-(626) 806-7574 http://www.ics.uci.edu/~rohit -- http://xent.ics.uci.edu/~FoRK