Whose is the text, Linus or Greg?

Holiday comes to us, it is harder to write and even read about planners and algorithms. Fortunately, there are more abstract topics in the world of open source software. Why not, for example, compare the style of business correspondence between Linus Torvalds and Greg Kroa-Hartman?

I suggest Habr's readers to practice physiognomy and ask themselves which of the two commanders of the Linux kernel project would you prefer to discuss the progress of your patch?

Warning , under the cat profanity in English, used in the context of business correspondence and scientific research.

If Linus Torvalds is heard all the time, the name of the second person in the kernel development project is much less likely to appear in news feeds. In the meantime, it is none other than the maintainer of a -stable kernel branch. He is the author of the linux-hotplug, udev and Linux Driver Project. He worked at Novell SUSE Linux, and since 2012 moved to the Linux Foundation . GKH fantastic performance and efficiency, the phrase maintainers don't scale has only a conditional relation to it.

Three scientists from the University of Elon in North Carolina, United States of America, conducted a study of the correspondence between Linus Torvalds and Greg Kroa-Hartman, taking the LKML archive for 20 years from 1995 to 2015. The goal was to study the correspondence styles of two project managers and learn how to distinguish their texts using the machine learning method.

For those who are interested in the topic, Linus Torvalds, being quite friendly and charming in real life, practices a rather tough and uncompromising style in business correspondence with the developers of the kernel. In connection with this circumstance, he repeatedly became the object of criticism from adherents of more stringent standards of business ethics.

There is no need to go far for examples. Recently, Kees Cook, a programmer from the Google Pixel team who tried to push security updates that could have negative consequences for software in user space, got the nuts. Fucking cretins (f ** ing morons) , as Linus described the security experts in the face of the persistent developer from Google.

Passion around SecureBoot and the rules of good tone

In July 2013, Matthew Garrett and Sarah Sharp left the project . To be honest, the loss was small and Matthew’s excessive willingness to meet the interests of large corporations repeatedly raised questions. Sarah herself in her blog demonstrates clearly not the style of communication that she wants to see towards herself.

I like it with “fart fart fart fart fart" © Sarah Sharp

Still, I'm not sure that even more thick-skinned developers would be able to withstand what Linus said to Matthew about the patches that should have provided Linux booting on workstations made under Windows 8+ and using UEFI with SecureBoot technology.

Guys, this is not a dick-sucking contest.

Further, the project leader expressed his point of view that such troughs under Microsoft are inappropriate on the part of kernel developers, and RedHat and other vendors are free to do what they like. In general, this correspondence was too expressive, even for LKML, and Garrett decided that he had enough.

In the course of this dispute, Linus challenged the GKH in the spirit of "get it together and stop being a rag," another of the key maintainers of Ingo Molnar said the same thing to him, arguing that sometimes developers can be sent to a known address.

Squabbles lasted for some time, and then the project participants agreed that a certain common set of rules for etiquette was needed, which was documented in /usr/src/linux/Documentation/CodeOfConflict

Authors of Linus and Greg

The study was conducted by the NLP method known as Naïve Bayes Classifier using the “bag-of-words”. The Python NLTK library was used to identify the parts of speech. A Monte Carlo simulation showed good prognostic abilities to establish text authorship using this machine learning algorithm. According to the authors of the study, the accuracy of the forecast reaches 97%.

The table below uses the profanity of the English language I think does not need comments.

The training table for the machine learning algorithm also speaks for itself.

The study takes into account only the vocabulary of the English language, which excludes elements of Finnish folklore, like this one.

I’m not talking about it.

As can be seen from the study, the creator of Linux is much more willing to use profanity and much less often is elegantly courteous, and yet I believe that the sense of proportion has not changed until now, and its rigidity has not crossed the line where self-indulgence begins. And what do the readers think about this?