Richard is to ask Frank Nagle to open the model/data for this research study.
Yes, please. I wrote up a response to the paper, and I reached out to Frank on LinkedIn and Twitter:
Why take the labor market approach over goods market? Buy, borrow, build. If “borrow” (OSS) disappears, isn’t it much more likely that companies would buy (goods market) than build (labor)?
In other words, here are the two basic alternate timelines for a world without OSS, using curl as an example:
- Build - Every company that uses curl—i.e., probably tens of thousands of companies—rewrites curl from scratch for themselves alone.
- Buy - One company (hopefully Daniel’s
) writes curl and sells it to every other company.
It seems obvious to me on its face that (2) is the realistic alternative worth exploring for purposes of a thought experiment to estimate the value of Open Source software. The paper includes (2) in an appendix, and it leads to a much different result:
- Build - Open Source is worth $9 trillion and companies would spend 3.5x on software without it.
- Buy - Open Source is worth $200 million and companies would spend 1.00005x on software without it.
Why did they report (1) when (2) is pretty clearly the more realistic answer?
Frank has not responded on social media. Manuel gave me a follow, so someone on the author team has noticed. I sent the following email just now:
Manuel, Frank, and Yanuo,
Greetings!
My name is Chad Whitacre. I’ve been working in the area of Open Source sustainability since 2012. I currently run the OSPO at Sentry, a software monitoring company, where I run a forward-thinking Open Source funding program. I recently read your working paper, “The Value of Open Source Software.” This is an important research topic, which I’ve taken a much less rigorous stab at in the past. Thank you for working on it.
As I was reading the body of the working paper, the main question that emerged for me was: why assume that all OSS “has to be recreated at each firm that uses a given piece of OSS” (12), rather than that commercial alternatives would become available? Then I read the appendix, which does in fact consider this scenario. It concludes:
Given all the complexities and assumptions one has to make due to lack of data for a goods-market approach, we place stronger emphasis on the labor cost approach highlighted in the main body but we include this method here for completeness.
I have to admit I find this unconvincing. It’s not obvious to me that the goods market approach displays any more complexity or requires any more assumptions than the labor market approach. The options for procuring software are buy, borrow, or build. In an alternate timeline without “borrow” (OSS), it seems so much more likely to me that “buy” (goods market) would prevail over “build” (labor market). In view of this likelihood, the goods market approach seems to me to deserve the bulk of the research effort, almost to the exclusion of the labor market approach.
Before submitting your working paper for peer review, it might be greatly strengthened by either a) making a thorough case for adopting the labor market over the goods market approach, or b) adopting the goods market approach as the primary focus.
Thank you again for your work on this important topic.