Sharpening the Axe: The Primacy of Toolmaking

Talk given at P99CONF 2022. Video: https://www.youtube.com/watch?v=_GpBkplsGus

Transcript

1. Brought to you by Sharpening the Axe: The Primacy of Toolmaking Bryan Cantrill CTO of Oxide Computer Company
2. OXIDE Give me six hours to chop down a tree and I will spend the first four sharpening the axe. – Abraham Lincoln Sharpening the axe
3. OXIDE Give me six hours to chop down a tree and I will spend the first four sharpening the axe. – Abraham Lincoln? Sharpening the axe?
4. OXIDE Give me six eight hours to chop down a tree and I will spend the first four six sharpening the axe. – Abraham Lincoln? Sharpening the axe?
5. OXIDE Give me six eight four hours to chop down a tree and I will spend the first four six two sharpening the axe. – Abraham Lincoln? Sharpening the axe?
6. OXIDE Give me six eight four an hours to chop down a tree and I will spend the first four six two forty-five minutes sharpening the axe. – Abraham Lincoln?! Sharpening the axe?
7. OXIDE Sharpening the axe?! The Home and Foreign Record of the Presbyterian Church in the United States of America, Volume 7, Number 1 (January 1856), Sermons for the Times: No. 2: The Dull Axe as cited in Quote Investigator, March 29, 2014
8. OXIDE “Wisdom is profitable to direct”? Translations for Ecclesiastes 10:10, from Biblehub.com
9. OXIDE A woodsman was once asked, “What would you do if you had just five minutes to chop down a tree?” He answered, “I would spend the first two and a half minutes sharpening my axe.” - C. R. Jaccarde Sharpening the axe! Objectives and Philosophy of Public Affairs Education, Increasing Understanding of Public Problems and Policies (1956), as cited in Quote Investigator, March 29, 2014
10. OXIDE Toolmaking • Better tools yield better artifacts • But the tools themselves are entirely invisible to the user – and are often not even present in the end product • Even though we know that toolmaking is important, we do not explicitly encourage it – it feels like a distraction • Many consequential developments in software history were in fact groups seeking to develop better tools for themselves…
11. OXIDE Unix, ca. 1969 The Evolution of the Unix Time-sharing System, AT&T Bell Laboratories Technical Journal, Vol. 63, No.8, (October 1984)
12. OXIDE TeX, ca.1977 Preliminary preliminary description of TeX, available as https://www.saildart.org/TEX1[1,DEK]1
13. OXIDE NSELite, ca. 1991 NSELite README, available as http://mcvoy.com/lm/nselite/README
14. OXIDE DTrace, ca. 2003 • System for dynamic instrumentation of production systems developed at Sun Microsystems, ~2001-2005 • Informed by 5+ years of trying to understand performance of systems • Put a bright light on the hidden innards of the system, exposing many (many!) glaring performance problems • Many of those problems were much further afield than we envisioned! • DTrace not only helped debug current issues, it greatly accelerated the development of software that came after it (e.g., ZFS)
15. OXIDE Jenkins (née Hudson), ca. 2004 Hudson, Kohsuke Kawaguchi as presented at JavaOne, 2007
16. OXIDE Rust, 2010 Project Servo: Technology from the past come to save the future from itself, Graydon Hoare as presented at the Mozilla Annual Summit, July 2010
17. OXIDE Toolmaking • These are just high profile examples – many more are never heard of • There are commonalities across all of these cases: ◦ Toolmakers were building for themselves ◦ The tools that they were building did not have immediate value on their own – their value was in delivering better artifacts ◦ Toolmakers retained the responsibility for the underlying artifacts • But why would toolmaking be discouraged?
18. OXIDE Discouraged toolmaking? • The challenge of toolmaking is that the payoff is often at a distance from the investment – both in terms of time and use • But the cost of toolmaking is immediate and clear: time spent making the tool is time not spent making the underlying artifact! • Toolmaking feels like it’s stealing time – it lurks in the shadows • But while it’s (in principle?) possible to spend too much time on tooling, it is unlikely to be the wrong use of time – just more difficult to quantify
19. OXIDE Toolmaking is an investment • With its clear short term costs and potential long term gains, toolmaking is best thought of as an investment • As with any investment, aggressive time horizons (e.g., days/weeks or small number of months) will result in overly conservative decisions • And like the best kind of investments, toolmaking also tends to be compounding: gains early in the lifetime result in more investment that yields yet more dividends • If toolmaking is always paying off, you may need to take more risk!
20. OXIDE Encouraging toolmaking • Yes, there is risk – but the downside is bounded to time spent • Indeed, the most common failure mode in engineering is not building too many tools, it is building something that no one wants • But in building a tool that one needs, at least one person wants it! • This is why toolmaking has in fact saved companies (e.g. Flickr, Slack, Docker): the tooling proved to have more interest than the thing! • When engineers wish to make tooling, they should be encouraged
21. OXIDE Toolmaking under fire • Encouraging toolmaking is hardest when deadlines are tight and opportunity cost seems high… • …but engineers themselves are cognizant of these constraints! • So when these constraints are present and engineers see a need for purpose-built tooling, that tooling should be prioritized • This is especially true for tools that improve the artifact by improving understanding: the best time to develop a debugger is when debugging!
22. OXIDE Trusting toolmakers • Toolmaking – and the organizational freedom that engineers need to engage in it – ultimately requires an essential ingredient: trust • In creative endeavors, trust must be the organizational binding force! • To encourage engineers to engage in toolmaking is saying: you are trusted to use your time wisely – and your future time is valued • To build an organization that values toolmaking, start by building trust
23. Brought to you by Bryan Cantrill bryan@oxidecomputer.com @bcantrill