Slide 1

Slide 1 text

On the abandonment and survival of open source projects: An empirical investigation ESEM 2019 Guilherme Avelino [email protected] UFPI/UFMG, Brazil Eleni Constantinou [email protected] UMONS, Belgium Marco Tulio Valente [email protected] UFMG, Brazil Alexander Serebrenik [email protected] TU/e, The Netherlands

Slide 2

Slide 2 text

2 Nearly all software today relies on open source code This type of code makes up the digital infrastructure of our society today Roads and Bridges, Nadia Eghbal

Slide 3

Slide 3 text

3 There is a growing concern on OSS sustainability

Slide 4

Slide 4 text

4

Slide 5

Slide 5 text

5 Millions of OSS repositories

Slide 6

Slide 6 text

6 Open Source Libraries

Slide 7

Slide 7 text

7 Who are maintaining the OSS projects we use?

Slide 8

Slide 8 text

8 How reliable are these communities?

Slide 9

Slide 9 text

9

Slide 10

Slide 10 text

Truck Factor The number of people on your team that have to be hit by a truck (or abandon) before the project is in serious trouble 10

Slide 11

Slide 11 text

ICPC 2016 11

Slide 12

Slide 12 text

Truck Factor Results 133 popular projects (TF ≤ 2) 65% 12

Slide 13

Slide 13 text

Discussions 13 "It would be useful to look at this when deciding on using a dependency in a project." "This is great avenue for research."

Slide 14

Slide 14 text

14 "It shouldn't be in trouble just because the current developers get hit by trucks. It's an open source project, anyone can start contributing if they want to."

Slide 15

Slide 15 text

On the abandonment and survival of open source projects

Slide 16

Slide 16 text

16 Research Questions RQ1: How common are TFDDs? RQ2: How often the projects survive TFDDs? RQ3: How surviving projects differ from non-surviving ones? TFDD = Truck Factor Developers Detachment

Slide 17

Slide 17 text

17 Dataset ▪ Top-500 in 6 languages ▪ Filtering step 1,932

Slide 18

Slide 18 text

18 Identifying TFDDs TFDD = all TF developers abandoned the system (more details in the paper)

Slide 19

Slide 19 text

19 Identifying TFDDs Jan/2015 Jan/2016 Today TF = 1 {Alice} ... Bob last commit ... TF = 2 {Alice, Bob} Alice last commit TFDD

Slide 20

Slide 20 text

20 RQ1. How common are TFDDs? 315 projects ▪ TFDDs: ▪ 66% => TF = 1 ▪ 59% => < 2 yrs of development

Slide 21

Slide 21 text

21 Identifying Surviving Projects Surviving project = new TF developer(s) assumed the project after TFDDs

Slide 22

Slide 22 text

22 Identifying Surviving Projects Jan/2012 Today ... TFDD ... TF = 2 {Alice, Bob} Jan/2017 TF = 2 {Bob, Charlotte} System Survived New TF developer

Slide 23

Slide 23 text

23 RQ2. Survival Rate 128 surviving projects ▪ Surviving projects ▪ 64% => one year after TFDD ▪ 48% => with newcomers

Slide 24

Slide 24 text

24 RQ3. Surviving vs Non-surviving

Slide 25

Slide 25 text

25 RQ3. Surviving vs Non-surviving Negligible or small effect size (Cliff's delta)

Slide 26

Slide 26 text

26 RQ3. Surviving vs Non-surviving Medium effect size (Cliff's delta)

Slide 27

Slide 27 text

27 Survey ▪ Target: new TF developers ▪ 140 emails sent ▪ 33 answers (response rate of 24%)

Slide 28

Slide 28 text

28 Discontinuation Risks 80% agree (or partially) agree

Slide 29

Slide 29 text

29 Motivations

Slide 30

Slide 30 text

30 Enablers and Barriers

Slide 31

Slide 31 text

31 Takeaways ▪ 16% of the projects face TF events ▪ 41% of the projects survive these events ▪ Key characteristic of surviving projects: friendly community

Slide 32

Slide 32 text

32 Replication Package http://doi.org/10.5281/zenodo.2546008

Slide 33

Slide 33 text

On the abandonment and survival of open source projects: An empirical investigation Thanks!