Operating Systems: Three Easy Pieces and the Quiet Case for Free Online Books

Most arguments for open educational resources arrive as policy abstractions. The story of Operating Systems: Three Easy Pieces (OSTEP) arrives as an accounting of what one professor did with his post-tenure spare time, and what happened next. That makes it more useful than the usual pitch, because the numbers and the mechanics are concrete.

Remzi Arpaci-Dusseau, who teaches CS 537, the undergraduate operating systems course at the University of Wisconsin-Madison, wrote the book with his wife and research collaborator Andrea Arpaci-Dusseau over many years. The chapters have been downloaded more than half a million times since 2012, and the project's web page drew close to three million views in a single year, boosted by the usual front-page traffic from Hacker News and Reddit. You can read the whole thing at ostep.org.

What's claimed

The headline claim is that a free online book, what the author calls a FOB, is simply better than a classic printed textbook. That is a strong statement, and it is worth separating the parts that hold up from the parts that are really just preferences.

The complaints that motivated the project are familiar to anyone who has bought a CS textbook. Price is the obvious one: paying north of a hundred dollars for a niche systems text while a celebrated novel costs ten dollars in softcover is hard to justify on production grounds. Edition churn is the second: frequent, low-substance revisions exist in part to break the used-book market, which would otherwise solve the price problem on its own. Quality is the third and most subjective: a handful of CS books are genuinely excellent, Hennessy and Patterson's Computer Architecture: A Quantitative Approach and Bryant and O'Hallaron's Computer Systems: A Programmer's Perspective among them, but a large middle tier is merely adequate.

What's actually new

The interesting part is not the critique, which has been made many times, but the process. The book started as plain text files written immediately after each lecture, with figures rendered in ASCII art. The author's address-space diagram, code at the top, heap growing down, stack growing up, free space in the middle, was the kind of crude sketch no publisher would print. Students did not care. They wanted reference material that matched the lectures, and the rough notes delivered that.

The rules that produced the draft are the transferable lesson here. Write plain text first, no typesetting. Skip real figures until the prose exists. Finish a note before starting the next one. LaTeX, proper diagrams, and homework problems came later, as incremental upgrades to something that already worked. This is the opposite of the way commercial textbooks are produced, where polish is front-loaded and the manuscript is held until it is complete.

The economics also defy the assumption that free kills revenue. Using on-demand printing through Lulu, the authors offer hardcover and softcover editions printed near cost, plus a full PDF with a glossary and systems-building tips for ten dollars. The PDF is the most popular paid format at roughly 55 percent of purchases, softcovers about 35 percent, hardcovers the rest. The margin is around five dollars a book, a deliberately small incentive rather than a business. The practical finding is that a meaningful fraction of readers will pay for a convenient bound or bundled version even when every chapter is free.

Why the model travels

Free chapters get used in ways printed books cannot. Instructors at schools across the U.S., Canada, India, Portugal, and Turkey have adopted OSTEP in place of the traditional choices. More telling is the casual traffic: someone searching for how a TLB works, what a condition variable does, or how file-system journaling is implemented can land on a single chapter without buying anything. That partial, search-driven usage is the real advantage. Content trapped inside an expensive volume is effectively invisible to a search engine and to anyone not already enrolled in a class.

The author lists the structural benefits plainly. Linkability means individual chapters can be cited and found. Broader readership follows from removing the purchase barrier. Partial usage lets an instructor assemble material from several free sources, which is cost-prohibitive with printed books. Frequent micro-revisions let the text improve continuously in response to reader feedback rather than on a publisher's release schedule.

Limitations and honest caveats

The account is refreshingly candid about confounds. The author credits the book with improving his teaching evaluations, then immediately notes that he was also simply getting better at teaching and putting more effort into the undergraduate course. The book was part of the improvement, not the whole of it. That kind of self-correction is rare in advocacy writing and makes the rest more credible.

There are real costs the model does not erase. Writing a book this way still takes years and depends on already having taught the material well enough to know how to explain it. The author was past tenure, with research funding secure, when he found the time. A junior faculty member chasing a tenure case has every incentive to spend that time on graduate courses and publications instead. Free online books lower the cost to readers; they do not lower the cost to authors, and the labor problem is exactly why good textbooks are scarce in the first place.

The claim that FOBs will tend to be higher quality over time is a hope, not a result. Continuous revision can improve a book, but only if a committed author keeps revising it. Most abandoned online drafts are evidence that the same low friction that lets a book ship early also lets it stall.

What it means for publishers

The author sketches a plausible surviving role for publishers: promotion, low-cost printing, and curation, while the content stays free online. Whether the industry adapts is left open, with a pessimistic nod to how slowly record labels, film studios, and academic journals have responded to similar pressure. His own conversations with publishers ended quickly once he insisted the text remain free chapter by chapter, which tells you how far the existing model is from the proposal.

The broader pattern is the one already visible across technical education. Course materials, lecture recordings, and reference documentation increasingly live on the open web because that is where they get read. OSTEP is not an argument that this should happen so much as a demonstration that it already works, complete with download curves, revenue splits, and a frank tally of what the author got wrong. For anyone weighing whether to put a manuscript online, the advice is unglamorous and probably correct: start writing, post what you write, use it in your own class, and let reader feedback do the rest.