CS papers need to stop publishing in the two-column format

Ever since receiving my Kindle for Xmas and uploading a PDF of a computer science paper I have realized that the way that CS papers are currently being formatted and published needs to change. In the world of CS academia papers are traditionally formatted in a two-column format. This format compared to a one-column format works out well as it fits more per page.

But is reading a CS conference paper on actual paper the future? At AOSD 2009 where I presented my paper I received the conference proceedings on a USB driver. Even though I had to go through pre-print, meet a stringent formatting guideline designed for print, and work with a pre-print service, I didn't get my paper in a physical format (I actually can't get an official hard copy of my paper so I am going to have to find a printer in Vancouver to do it so I can have an archival copy for myself). If we are not even going to receive a paper copy why worry about how it will look on the printed page?

And this brings up to the second problem of how two-column PDFs look on a digital screen. If you read a two-column PDF on an LCD it is a real pain because you typically have zoomed in enough to make the text readable that an entire column will not fit on the screen at any one time. That means to read a single page you must start at the top of the first column, scroll down to finish that column, go back up the page to start the second column, and then scroll down again to finish that second column. As more papers are read on computers it makes less and less sense to format papers this way.

But what really makes the two-column format bad is PDFs inability to reflow its text on e-readers. If you put a PDF on an e-reader that is not the actual size of a sheet of paper the PDF must either be resized to fit entirely on the screen or software has to try to figure out how to reflow the text which can lead to bad results. On my Kindle I have to turn to landscape mode in order to read a two-column PDF, but that simply zooms enough to fit the width of the page on the reader which still leads to small text.

If CS papers switched to a one-column format these issues of reading on electronic devices would simply go away. PDFs would be readable on a computer screen as you would simply scroll in one direction: down. As for e-readers, I think honestly the best solution in all of this is to write text with reflow in mind and then have a standard way of converting from EPUB to PDF. This would be easy if people simply didn't embed table, charts, etc. directly in the paper but instead put them in an appendix and had links to the images. But even if they are inline, making papers simply one-column goes a long way as long as no one is doing crazy typography (which is typically frowned upon in conference papers thanks to length restrictions).

Luckily I am not the only person thinking about this. The journal Cell has begun developing a new paper format. While currently aimed at displaying better in the browser, that requirement should lead to easy packaging in EPUB for e-readers.