Behind the Scenes of A Fire Upon the Deep
by Vernor Vinge

A Primitive Form of Story Documentation

Since 1979 I've used the convention that lines beginning with "^" should not normally be printed. This makes it easy for me to "comment my text". Over the years, as storage capacities increased, I found that even this extremely crude tactic could be very helpful in story development. About one fourth of my Fire Upon the Deep manuscript is such hidden commentary. These comments served a variety of purposes, and I used various tag words to discriminate between these purposes (see the table below). Besides formal tags, I had a large number of key words to identify different aspects of the story. I used various tools -- mainly grep -- to follow the key words around the manuscript. Note that this technique is not hypertext (... well, maybe it could be called a "manual form" of hypertext, with grep being used to dynamically compute links :-).
Now Brad Templeton of Clarinet has taken these commented files and put them into a simple hypertext format based on my standard tags. In this form, it is possible to read the story without being bothered by the comments -- yet be able to see the comments on demand. (Because of production deadlines I have not seen the exact user interface for the Clarinet edition, and so some of this discussion may be slightly inconsistent with details of the final product.)
I'm quite excited about what Clarinet has done with the manuscript. I think this edition will be fun for people who want to look behind the scenes at a story as it is being constructed and discussed. In preparing the manuscript for Clarinet, I have tried to extend and clarify the notes. However, I want to warn you that since these are mainly internal development notes, they are often cryptic, repetitive, and inconsistent. The arithmetic has not been rechecked! (The notes are also tentative in the sense that they may be contradicted by the later-written sequels and prequels.) There are interesting things in these notes, but you can get lost in the tedium of minor issues or be led astray by discussion of problems that were later solved (leaving the discussion without referent!). Hopefully, this overview will make it more convenient for people who do want to look at the notes.


Editor's Note: Read the file exnotes.rtf to read details of the internal formatter notation used by the author. This notation was converted by our software to the final document you received, so it is not visible to you.

I use tags a lot. In a sense almost anything could become a tag (and a target for grep), but the most formalized tags and their meanings are as follows. (Clarinet has associated these tags with hypertext icons.)
AWK The referenced prose seems awkward. (Hopefully most of these are fixed.)
CHK The comment involves something that I should verify.
CHRON A timestamp on the writing or reworking of this area.
DONE The action suggested in the comment has been done.
FRAG The commented text is a fragment -- either something that is looking for a home in the manuscript, or perhaps the reverse, namely text that has been removed/replaced. In the latter case, the fragment may contradict the present story line.
ID The comment is an idea related to this story.
IDEA The comment is an idea unrelated to this story.
IMP The comment was thought to be important.
INCON A possible inconsistency.
IMPER The comment was thought to be highly important.
MARK The comment was a book mark used during traversals.
NO I didn't follow through on the suggestion in the comment.
NÆH I decided there was no need to follow through on the suggestion.
PRB The comment relates to a problem about the story.
PRO The comment describes changes that should be made later in the story ("prowritten").
QU The comment is a question that I have.
REN The comment relates to the names in the story (or naming issues).
RETRO The comment describes changes that should be made earlier in the story ("retrowritten").
SEQ The comment is about a sequel (or prequel) possibility.
SOL The comment discusses a solution to some difficulty I have.
SOLN The comment discusses a solution to some difficulty I have.
SORRY I couldn't follow through on the suggestion in the comment.
TITLE The comment is a title suggestion.
TUF I didn't follow through on the suggestion in the comment.
If the first letter of a tag is lower case this means the issue in question has been taken care of. (However, this convention is not consistently followed.) CHKd means I did verify the information in question.
Most unattributed comments are by me. Comments from consultants are normally prefixed by the consultant's initials. (In the Clarinet edition a red letter "V" is used in place of my own initials, "vsv". The pronouns "I" and "you" often refer to vsv; I talk to myself a lot, with and without special notations!)
As my work progressed, there came to be layers of comments. In most cases these can be distinguished by the added tags (DONE, SOL, ...). In other cases, you can see a dialog between myself and my consultants. At one point (Xmas Break 1990?), I had numerous pending issues in my "^ " comments. I swept through them, inserting "!!" as I processed. Thus the stuff after "!!" is normally the old comment together with whatever action I took. (In the Clarinet edition, I believe that Brad Templeton has mapped the "!!" to a single red "!".)
Some of the notes have dates attached to them. These dates are accurate for some part of the note they appear in. Unfortunately, they don't reflect everything about the note -- and many comments have no date attached to them. (There is another style of date you will notice in some places, dates like 5 June 15989! This was my crude way of keeping track of the story chronology. Since Tines' World days are about the same length as Earth's, it was possible for me to use the calendar tools in Lotus Agenda(tm) to manage the chronology of the story. I more or less arbitrarily set the beginning of the story (the birth of the Straumli Blight) at 00:00:00 23 June 15988. Note that this assumption never leaks out of the annotations; it was just an artifice to make it easier for me to track the relative position of events occurring within the story. (In fact, it's not certain just how far in the future this story happens; fourteen thousand years after 1988 seems to be about the minimum.)
I used a few unusual characters in the manuscript (mainly in the annotations -- I took pity on myself and the copyeditors). These characters are listed in the following table. Hopefully, this table will be helpful if your display device is not mapping these 8-bit patterns to the characters I intended!
Editor's Note: We have converted these special characters to their RTF equivalents in the ANSI RTF character set. To read about internal encodings, read the file vvencoding.


A Little History
Some of the comments may be more intelligible if I say a little about the history of the writing of A Fire Upon the Deep.
During Xmas Break 1987/88, I wrote a novella, "The Blabber" (later published in my story collection "Threats and Other Promises", Baen Books, 1988). In fact, this story was a sequel to A Fire Upon the Deep. The novella was written after most of the background universe of the Zones and the Tines had been worked out, but before I knew the details of the novel. It might seem that the existence of such a story would make the planning of the novel more difficult. To some extent this is true, but more often "The Blabber" provided boundary conditions and inspiration for the novel. And the apparently unresolved inconsistencies between the novella and the novel may provide fuel for later revelations. Some of the notes in the novel reference "The Blabber" and these consistency issues.
I was on leave during the 1988-1989 academic year. One of my projects during that time was to get a start on the novel. During the summer of 1988 I attended the distributed systems school at the University of Tromsø, and I visited Oslo. I saw much that affected the story (and I met many wonderful people -- see the book's acknowledgments). During the following twelve months I completed an initial draft of most of the story. In programming there is a saying that "First you write the program; then you do it again, and this time you do it right." My experience was not quite this discouraging. The Tines' World plot line went smoothly right from the start, but the Ravna plot line was enormously difficult. In the earliest draft, Ravna was a secret agent opposed to the Straumli Realm, an empire that had unleashed the Blight in a desperate effort to avoid losing a war. In the wake of this catastrophe, she escapes to Relay where she engages a Vrinimi humanoid, Yrdnalf Scrubscooey, to help her find the refugees from the High Lab. I pursued this version through the Fall of Relay before I realized that it was unacceptably weak. It was time for me to retreat, salvage what I could, and plan a different approach: Having Ravna be an employee of Vrinimi made the background much easier to present. Having Pham be ... what he is ... gave me a direct witness on the Slow Zone and the Depths and the nature of the Powers. (And of course, it provided a romance that was within my ability to describe.) Of course it also meant that I had to completely rewrite the Relay scenes. In the comments, you will find only fleeting references to poor Yrdnalf Scrubscooey and Straum the warlike realm; now they are less than fiction.
The only other major change was small by comparison, but it also illustrates my problem with the Ravna plotline. It became clear that the flight of the Out of Band II must take many months, yet I had only a few scenes aboard the ship. I had not properly mastered "summing up". I call these gaps "Lost Times" in the notes. Filling them in with retrospective summaries turned out to be easy as well as necessary -- and gave me an opportunity to improve the characters and the credibility of the plot. Some of this strategy is discussed in commentary tagged to "Lost Time".
Finally, when Brad Templeton suggested (April 1993) that I make the annotations available for the Clarinet edition, I looked at the notes and marveled at how cryptic and contradictory they were. I went over them and tried to make them more intelligible. In places, I added some background about terminology (see "evocation" and "virtual partition"). I also added my thoughts about a few things that readers had commented on.

The Future of Annotations
You've probably noticed from the discussion above that my word-processing automation is very primitive. I like the fact that written fiction is built on words and sentences and paragraphs. There may be times when fonts and page layout are important to the story being told, but they are rarely at the heart of the prose art form.
Software for managing story development is a different matter:
My embedded comment convention served me well during the 1980s. Early in the decade, disk capacities were so small that the comments served mainly for important reminders. By the end of the decade, I could keep an entire novel in a single file (if I wished), and I could keep much of the support documentation in-line with the text. Fortunately, processing power had also improved, and by using grep I could hop around the manuscript, guided by whatever tags were useful at the moment.
For a story the size of A Fire Upon the Deep, this was not enough. I needed more than the ability to find tags and instances; I needed to generate summaries of background information and story text, keyed on the parameters of the moment. Late in the project, I began to use Lotus Agenda(tm). (Stan Schmidt describes a similar experience with Hypercard(tm) in "Hypertext as a Writing Tool", SFFWA Bulletin, Summer 1992, p6-10.) Agenda was very helpful, especially in tying a consistent timeline to the story events. My experience with Agenda convinced me that what I wanted was possible and could be as useful as I hoped. Unfortunately, home computers in 1990 were too small to support everything I needed. I needed a system capable of supporting all the story notes and the story text in one multilinked structure.
There are two other features that would have been useful -- though both are such an infringement on privacy and ego that I find them a bit scarey:
Whenever I wanted my consultants to look at the manuscript, I had to print up a hardcopy and ship it to those consultants. Even if I had sent the files over the phone, the operation would still have been tedious. Much nicer (and now possible with the Internet) would be software that makes a project visible (to the extent that the author desires) to others and allows them to enter comment nodes and pointers. You see a small amount of such dialog in A Fire Upon the Deep but that is the result of much manual labor on my part. In 1992(?), Barry R. Levin had an article in Publishers Weekly, "Manuscript Collection -- An Endangered Species". (I have not had a chance to read his article myself. What I say about it is based on Stan Schmidt's description in "The Manuscript that Never Was" (editorial), Analog, February 1993, p4-12.) Barry Levin notes that because of word processors, collectors and academics have lost virtually all their former access to the writing process (namely access to marked-up rough drafts). I fear Mr. Levin is correct about the problem of collectors and "one of a kind" physical manuscripts. But there are some happy possibilities for literary researchers: In addition to the tools I describe above, one can imagine word processors that store not just the author's text but also a timestamped record of every keypress. Of course, this would be done transparently; the author could completely ignore this feature. The resulting files would be much larger than simple text files. Nevertheless, the storage requirements are well within the capabilities of 1993 home computers. (In fact, there are existing editors -- such as emacs -- which could be modified rather easily to provide this service. It is a much easier trick than hypertext and groupware facilities.) Imagine: With a proper "viewer" program, literary researchers could watch every keystroke (even backspaces and paragraph deletes) with the exact timing of the original! It would almost be like having a camera in the room with the author, watching him/her work. What a window on the creative process -- the long pauses indicating just where the the Great Author needed deep concentration (or perhaps where the Great Author needed to make a trip to the bathroom). Of course most authors would forthrightly reject this infringement on their privacy. But perhaps there is intermediate ground: The "spy word processor" might allow the author to mask and alter the more intimate details of his/her life. My guess is that many authors would be willing participants in such an enterprise (and for the academics it would still be a substantial improvement over the present levels of access).
The software tools I've described all exist in some form right now. There has been much genius applied to such things (eg, Ted Nelson's ideas and the work of the Xanadu people). There are academic and industry teams working on super versions of groupware and hypertext appropriate to the super hardware that we will have in the near future.
When tools of great power are invented, an interesting thing often happens: The original problem that the tools were invented for suddenly seems less important, and a whole new field of endeavor is created: "Invention is the mother of necessity" (not my quote, but I don't know the source). This is exactly the situation with groupware and hypertext and multimedia tools. Notice that all my discussion of such things has been in support of classical, written fiction. What you have in the Clarinet hypertext edition of A Fire Upon the Deep is simply the working support for a conventional novel. It is not to be confused with the idea of hypertext fiction. I believe hypertext fiction will ultimately be an entirely new art form, as different from novels as motion pictures are from oil paintings. (We don't have many prototypes of this yet. In some sense, computerized adventure games are an example. Marc Stiegler's pioneering hypertext version of his novel David's Sling is a more direct example.) I believe that some of the most significant aspects of the new medium are yet to be recognized. Guessing: There may not be hypertext sequels so much as the instantiation of new windows on the "reality" of the story. Group participation both during initial construction and in expanding the ongoing reality may be one of the most striking features of the art form. (See Chip Morningstar and F. Randall Farmer, "The Lessons of Lucasfilm's Habitat", in Cyberspace: First Steps, edited by Michael Benedikt, MIT Press, 1992.) Hypertext fiction may evolve into immense art works that combine the essence of professional production teams with independent artists with the interests and efforts of the ultimate viewers.
All this is awesome, wonderful stuff. And the skills needed for success in true hypertext fiction are very different from those needed to write a conventional novel. But I believe that the efforts of "classical" novelists will continue to find acceptance in the marketplace, even as those novelists use more and more advanced tools to support their craft. Just as the art of painting in oils still flourishes alongside cinema, so there will be a place for novels and novelists.