What is it that is the same, and what is different about a totally electronic journal? We detail several of the elements of an ordinary journal and show how the proposed journal expands all of these features.

Table of Contents: Each article will be identified by author and title.

The author's name is a link to the author's photograph and a paragraph of biography.

These are linked to the author's home page on the World Wide Web, where the reader can find a list of the author's bibliography as well as a curriculum vitae.

The title is a link to the abstract and to the keywords and subject classifications as well as to the article itself.

Abstract: There will be at least two levels of abstract, one short and one extended. Basically it should be possible to read each article at at least two levels, one for the experts, one for the general mathematical reader. The abstracts should indicate the required background for each version.

Keywords: These should place the article in context, showing related areas. It should be possible to get to the Mathematical Reviews classification scheme, to see the primary and secondary classifications.

Glossary: One of the ways in which the readership can be expanded is by providing glossary entries for various concepts. These can be chosen from a standard collection of "encyclopedia entries" or written specifically for an article. These too can be linked to one another. This way, the article itself does not have to define standard terms within the text, although specific terms should be defined in the text itself.

Background Material: If the article extends earlier results, it may be possible to summarize the background, or link to other published results. Even textbook material might be accessible on line.

Examples: Each time a concept is introduced, there can be a set of examples giving the reader a better idea of what it means. The examples can be graduated, from the simple to the more complicated. The illustrations can be static, or dynamic, with mpeg movies, or video footage. There can also be a link to an interactive program such as Fnord, or Maple or Mathematica.

Responses: It should be possible for readers to respond to the author, mentioning related results or even suggesting ways of proving the results in other ways. There might even be "exercises" in the text, encouraging readers to reply (the way Martin Gardner used to put questions in his Scientific American column.) Example--my two articles on triple points of immersed surfaces generated a number of responses of the type, "I know another proof of that result." That is not the same as, "I know where there is a published proof of that result."

It would also be possible to handle responses in the form of a moderated newsgroup, so other readers could profit from a discussion of the topic.

Future Modifications of Articles: It will be possible to include forward referencing links so that any reader can find places that refer to the article at hand. In particular, if there are any errata, there can be links in the article to indicate them. For archival purposes, the original article should probably be preserved intact, even though a later reader might choose to look at a corrected version. This could be a bit disquieting to an author, to have the uncorrected original always sitting there, but that is what happens if it is published in paper. The difference is that someone reading the original might not know the embarrassing fact that the paper required correcting.

Staffing: For the original setup of the journal, we should require half-time support of a skilled person at the level of a post-doc at the Geometry Center, plus the part-time work of one or two graduate students. There will have to be standard volunteer editorial work to certify content. The idea would be to develop models that could then be used by other mathematicians or mathematical support staff at other institutions, so that later on less effort by Ph. D. level individuals would be required.