Please read these guidelines carefully. They were written to help you give your formal submission its best possible chance to be accepted.

WHAT KINDS OF PAPERS DOES USENIX PUBLISH?
The most important thought to keep in mind when deciding whether to submit a paper is "what will the audience or readers learn from my paper?" We don't expect every paper to report on a major breakthrough, but we do look for something new, potentially useful, and not entirely obvious. Think about how different your work is from previously published papers; it may be good work but if there is nothing new to learn, it isn't worth reading (or writing) a paper about it. Think about how other people might find your work useful; can they apply what you are teaching them to their own systems? And, does your work really improve upon the previous state of the art? Or does it show how other people have been confused? "Negative results" that contradict the conventional wisdom are often more important than positive results.

Trying to decide if something is non-obvious isn't easy (patent lawyers make lots of money arguing about this), and sometimes the best ideas seem obvious in hindsight; but if lots of people have done the same thing, and you are simply the first person to have considered writing a paper about it, perhaps it's too obvious.

Also think about whether this conference is the right place to publish your paper. Perhaps it belongs in the USENIX annual technical conference, a more theoretical conference, or a conference with a different kind of focus. Or perhaps it doesn't belong in a conference at all; it might be more appropriate for the USENIX newsletter ;login:. On the other hand, USENIX conferences typically cover a broad range of practical issues in open computing systems. It may be that you can take a paper that has several possible themes, and write it to concentrate on issues specifically interesting to a USENIX audience.

The program committee will also be trying to decide if papers will lead to a good 25-minute presentation. Some systems are just too complex to be presented this way (perhaps you should focus on just one aspect); other papers just don't have enough to talk about for that long. On the other hand, a few rare papers are accepted mostly because the committee expects them to produce an interesting talk, but that might not otherwise merit publication.

Again, when you are writing your paper, keep in mind "what do I intend to teach the reader?" That means keeping the paper focused on one or a few main points. Don't try to cram too many big issues into the paper, and don't fill it up with irrelevant details. But do include enough background for the reader to understand why your problem is important, how your work relates to previous work in the field, and how it might fit into a practical system. Also, provide enough detail for the reader to put your performance measurements in context. It is vitally important to provide a good bibliography, both so that you give proper credit to previous work, and so that a reader can know where to turn to find additional background information. The program committee will not look kindly on a paper if the author doesn't appear to be familiar with the current literature.

MORE INFORMATION IS AVAILABLE
Lots of papers and books have been written about how to write a good paper. We'd like to suggest that you read a paper called An Evaluation of the Ninth SOSP Submissions; or, How (and How Not) to Write a Good Systems Paper. This was written by Roy Levin and David D. Redell, the program committee co-chairs for SOSP-9, and first appeared in ACM SIGOPS Operating Systems Review, Vol. 17, No. 3 (July, 1983), pages 35-40.

Although SOSP and USENIX papers do differ somewhat, Levin and Redell give good advice for authors of any kind of systems paper.

Another helpful paper is: "The Science of Scientific Writing", George D. Gopen and Judith A. Swan, American Scientist, Vol. 78, No. 6 (Nov-Dec, 1990), pp. 550-558.

This article describes not how to write an entire paper, but how to write sentences and paragraphs that readers can understand. Unfortunately, due to copyright restrictions we cannot make this available online or send you photocopies, but almost any library should have copies of this magazine.

We also recommend that you read the proceedings of some recent USENIX conferences to get an idea of what kinds of papers are published. Not every one of these papers is perfect (or even great), but most of them are better than most of the ones that got rejected.

HOW SHOULD I GET MY MANUSCRIPT TO YOU?
The Program Committee would prefer to receive submissions via the Web form located on the conference Web site:

• The web form will also require the authors to provide the following information:
  • The title and authors of the manuscript.
  • The name of one author who will serve as a contact, with regular and electronic mail addresses, daytime and evening telephone numbers, and a fax number.
  • An indication of which, if any, of the authors are full-time students.

• All submissions will be acknowledged.

• Prizes will be awarded for Best Paper and Best Student Paper.

PostScript files are also acceptable, but given that PostScript generators vary quite a lot, it is likely that we may not be able to print every submitted PostScript file. For instance, several software packages generate PostScript that can only be printed on Apple Laserwriters. So if you submit PostScript documents, please remember the following:

• Use only the most basic of fonts (TimesRoman, Helvetica, Courier). Other fonts are often not available with every printer or previewer.
• PostScript that requires some special prolog to be loaded into the printer probably will not work for us. Please don't send such files!
• If you used a PC or Macintosh-based word processor to generate your PostScript, ensure that it can be printed it on a more generic PostScript printer before sending it so as to ensure that the PostScript is portable.

Good Luck!