The core basics of design are to know for whom you design, that is who are the users of your solution, what they expect and what they need. It is also necessary to assess if the actual user of your product is the one you designed it for in the first place, that is, who is missing from your user base, to avoid the survivor bias .

This is unfortunately overlooked in FLOSS and it is often not well regarded to collect user data. But then, developers will justify their design by referring to the “average user” as a fantasized entity, undefined and certainly not understood.

In 2020, I ran a survey of the darktable user base to gather actual data. Since these past 2 years have been hectic, dev-wise, I only got to process the data this summer.

#Methodology

#Survey

The survey was conducted in English, on the internet, using poll-maker.com  to record answers. It had a total of 26 questions. The link to the survey was posted :

It was relayed :

It is to be noted that it was not relayed on Github or any other code hosting platform, though the IRC channel is populated with developers and power-users, so the respondents are expected to be mostly users and not contributors.

1101 respondents from 59 countries participated between March 6th 2020 and March 20th 2020. 1085 respondents answered all questions. Between 2019 and 2021, 66 persons have contributed to darktable source code , so we can expect that at most 6 % of the respondents are developers and contributors of the project.

The polling system (poll-maker.com) recorded IP addresses and user-agent to prevent people from taking the survey more than once.

#Software stack

The raw data has been exported from poll-maker.com in CSV and post-processed in Python with Pandas , Numpy  and the graphs are rendered using Plotly . The Python code written for this task has been turned into a library of functions that can be reused in the future.

The color scales used in the graphs have been designed for perceptual uniformity using the darktable UCS 2022 color space Python code and Colour .

This article is rendered from Markdown and Plotly HTML exports on WordPress using my own plugin : WP Scholar.

#Cost

The survey cost me a monthly subscription to poll-maker.com, that is around 40 $ out of my own pocket.

#Results

#User’s origin

The countries were recorded based on the IP addresses and may be wrong if a VPN was used. The results from 59 countries were recorded, but only 35 countries had at least 3 respondents.

image
IP adresses heatmap

Number of respondents by country

We see a predominance of Western Europe as well as North America. It is worth noting that most of the development team was German before 2018, and French between 2018 and 2020, and the proximity between developers and users may be an explanation for the massive engagement in the survey of those communities.

Number of respondents by country

The survey being written in English, the language barrier is expected to bias the results by favouring highly educated and slightly younger people in countries where English is not the official language. Therefore, the rest of this study will systematically split the respondents assumed to be native English speakers (whose IP came from the USA, UK, Canada, Australia, New-Zealand and Ireland) and the others, assumed to be non-native speakers, as two different samples. South Africa was not retained as a native English country, since the English speakers account for 9.6% of the population . This gives us 308 native speakers, and 793 non-native.

#User’s identity

#Age

As expected, the distribution of ages is flatter for native than for non-native English speakers, which is is more likely linked to education and language. Interestingly, in both samples, the most represented class is the $[36;45]$.

For both groups, the median is within the class $[46;55]$. The range $[26; 65]$ contains 81 % of the native and 85 % of the non-native respondents.

Then, I want to assess if the population of darktable users is a random subset of the population of digital photographers, or if it is non-randomly biased by some effect. But since there are no such figures for photographers, the next best metric is the general population of internet users. To do so, we need to check if the age distribution matches the age distribution of computer users.

Using the USA population as a control sample, I took the age distribution in the general population in early 2020 from the United States Census Bureau , cut it into age classes, and estimated the number of internet users from the 2021 ratios given by the Pew Research Center . The ratio of internet users is between 99 and 96% for all age classes, except for the 65+ where it drops to 75%.

I also used here the age distribution of the viewers of my Youtube channel , dedicated to photo editing with darktable since 2016. The statistics used here cover the years 2019, 2020, 2021 in full, which include 309 210 views. This is to be taken with precaution though, because my videos are sometimes posted in French, sometimes in English, and they tend to dive deeper into the technicalities than the average. In any case, it might be interesting to see how they compare.

It is immediately clear that the US population of internet users has a flat distribution and that both the survey and Youtube show a an abnormal peak between 40 and 50. By comparison, the folks below 35 and above 65 are under-represented in the survey, but everybody above 35 is over-represented on my channel.

The population of darktable users is therefore a non-randomly filtered subset of the general population of computer users, which clearly indicates darktable fails to reach or to appeal to younger folks. This should be compared with the age distribution of other imaging software to conclude, many factors could explain it :

  • younger folks use their smartphone and mobile apps to take and edit photos,
  • darktable is meant to process raw photographs, mostly coming from dedicated cameras (high-end point-and-shoot, DSLR, mirrorless) costing at least 800 €, which is too expensive for younger folk (or may be a low-priority expense given what smartphones can achieve now).

#Gender

I did not want to engage in the americano-centric complicated gender problems of the time and therefore did not poll the gender of respondents.

There is no woman in the darktable development team, and that I have met only 4 or 5 of them on the various darktable forums and chats, in both French and English. On my channel, Youtube reports that 0.3 to 1.3% of the viewers are women, depending on years (0.7% on average), and Bruce Williams  confirmed to me he gets figures below 1% too.

The usual explanation to this is that women are less interested in these technical topics, but it doesn’t hold. Indeed, since the 1980’s, in Western countries, 30 to 40 % of computer science degrees were earned by women, and they account for 20 to 30 % of the programming workforce in software and video games companies. In engineering, 15 to 25 % of the degrees have been earned by women in the same period, and they account now for 20 to 25 % of the workforce in engineering companies. So women are interested in technical matters, they are a quarter of the technical staff, and we would expect at least 2 of them among the developers.

But it doesn’t stop there. The US Bureau of Labor Statistics  reports that 49.3 % of professional US photographers are women in 2021 and women account for 75% of the photography graduates in the USA . In France , women account for 63 % of photography graduates, 60 % of employees in photography companies and 37 % of the self-employed professional photographers. So, in both countries, it’s at least half of the staff and way more than half the graduates, and it has been so for at least the past decade.

Yet we have less than 1% of them among darktable users. This definitely points toward non-random filtering that cast women away from darktable.

It is all the more surprising that a French report from 2019 by Marie Docher , commissioned by the Ministry of Culture, mentions that female photographers earn a median income of 1000 €, compared to 1400 € for males (all of them combining anyway, on average, 3.5 jobs to join both ends…). darktable being absolutely free to use should appeal to them even more, at least more than the 12€/month of the Adobe Cloud subscription. And yet it does not.

#Education

The education part is tricky since the curriculum varies greatly between countries : types of schools, number of years to obtain each degree, etc. are not uniform and difficult to translate. I chose to record education in terms of years, with no distinction between the actual type of school attended (noticeably, Germany has 3 kinds of high-schools — Hauptschule, Realschule, Gymnasium – which have different opportunities at the end) or degree earned (bachelor or associate degree in US colleges, for example).

These results confirm the hypothesis of the non-native English speakers biasing the results : highly educated people are more likely to have a proficient English level and therefore to take the survey. Nearly half of the native group (46 %) have a college degree, and more than half of the non-native group (51 %) have a master’s or a PhD.

To check for the education bias among the US respondents only, I want to redo the same comparison of the distribution against the education level of the general population. The distribution of degrees among the US population aged 25 or more is taken from the US Bureau of Census for the year 2021 . Consequently, the age range $[14; 25]$ is removed from the US users in the survey, leaving a sample of 146 US users.

We see here a bias toward the most educated people:

  • the part of darktable US users having at least a master’s is more than twice (35 %) the part of the general population (14 %),
  • the part of darktable US users having less than college education is less than half (15 %) the part of the general population (37 %),
  • the part of users having college education is very close to the part of the general population (50 %).

Looking at the heat map of the respondent above, this is consistent with the geographic origin of the US respondents, located on the East Coast, California, Washington and Oregon where all the prestigious colleges are located.

#Studying background

The background is the field in which the respondents studied. The survey accepted more than one answer per person, and 1227 responses were given from 1101 respondents. The choices were :

  • Graphic arts (painting, photography, drawing…),
  • Other arts (dance, music, theatre, architecture…),
  • Technics (craftsmanships: cooking, construction, plumbing…),
  • Applied and fundamental sciences (maths, physics, biology, engineering, health sciences…)
  • Humanities (literature, philosophy, sociology, psychology),
  • Business & Administration, Finance,
  • Law,
  • Military,
  • I did not study past elementary school

Almost 61% of the non-native and 63 % of the native speakers have a scientific background, while only 6 % of the non-native and 10 % of the native speakers have a graphic arts background. The distribution is close for both native and non-native, with marginal differences for technics/craftmanships and business/administration/finance. The law and military fields have each 15 respondents over both groups, they are therefore to be taken with caution.

It is to be noted that the workforce in visual and performing arts was estimated at 2.12 M  in the USA in 2020, while science, engineering and computer science had a workforce estimated at 9.48 M  the same year. Among the native speakers, we find an art vs. science ratio of 23/100, while the general population has a ratio of 22/100. What remains to be seen is how much both fields weigh in the population of photographers, as there is no guaranty that the population of photographers is an uniform subset of the general population.

Next, I want to assess if we can find user clusters on a 2D graph showing education level and studying background. To do this, I will create a density heat map for both native and non-native speakers. The numbers are the densities, given in percent of the number of respondents.

These graphs confirm that, for both groups (native and non-native), 47 to 50 % of the user base has a bachelor or a master’s of sciences, which corresponds roughly to an engineer level depending on countries. It is to be noted that the engineer level matches a bachelor of sciences in North-America, while it matches a master’s of sciences in Europe (but only since the 2000’s), which may explain the offset between the native and non-native groups.

Both groups have also 11% of PhD of sciences, which is more than the total of graphic arts (6 and 10%) for all education levels in both groups.

Then, I want to redo the same study with regard to the age distribution:

We see that the bulk of users with sciences and business background is 10 years older for native speakers than for non-native, which could be linked to the fact that the people proficient in English in non-English-speaking countries are more likely young. We see that 48 % of the non-native group are people between 26 and 55 with a scientific background, and 54 % of the native group are people between 26 and 65 with a scientific background.

#Conclusion on user’s identity

From all the elements above, we can summarize that :

  • darktable’s users are more educated than the average population, with twice as much university-educated persons,
  • darktable’s users distribution is biased toward middle-aged people, with 13.6% more people in the $[36;65]$ range and 13.3% less people in the $[14; 35]$ range, compared to the general population of internet users,
  • darktable’s users are biased toward scientists, with 61 to 63 % of users having a scientific background,
  • around half of darktable’s users have an engineer level or equivalent (bachelor or master’s of science),
  • there are more PhD in sciences (11 %) than graphic art graduates (10 %) among darktable users,
  • there are close to no women.

The archetypal darktable user is a 50-years-old male, [-25; + 15] years, with an engineering degree or equivalent. This archetype matches 55% of the user base.

It is very homogenous social group, which does not represent the variety of photographers, and favours highly educated and technical/scientific profiles. It is concerning for a photo editing application to attract an absolute majority of scientists and an anecdotal minority of art graduates.

#User’s computing skills

One of the main criticisms toward darktable is its general user-experience is unwelcoming to non-engineers and too complicated to most computer users. I want here to assess here if the data is biased in a way that verifies this assertion.

#Ability to interact with the computer

To evaluate the ability to interact with the computers, I asked a question about typical tasks that the respondents could achieve. Respondents could give only one answer to the question “How comfortable are you with computers ?” and the possible answers were :

  • Not much. Someone handles the computer administration for me. Level 0. Basic user with no admin skills,
  • I know how to install/update software, but I need help when error messages appear. Level 1. Basic user able to run updates,
  • I know how to solve errors as long as no console command line is required. Level 2. Basic user with admin skills,
  • I know how to use the shell and command line to access system configuration. Level 3. Power user,
  • I know how to loop over files in shell to apply batch operations. Level 4. High-level programmer,
  • I know the output to ( (int)x << 1 == (int)x * 2 ). Note : x << 1 is a bitshift operation in C syntax. Level 5. Low-level programmer.
Note

In computer sciences, the ground level is the hardware. Software is applied on top, but can be closer (lower) or farther (higher) from the hardware level.

“Low level” programming uses languages that are not developer-friendly and deal directly with the hardware layer. “High level” programming uses scripting languages that abstract the hardware concerns (like memory layouts, protocols, etc.).

Both ways of programming share the same logical structures. However, the low-level way require more skills from the programmers, while the high-level way enables people without formal computer science/electrical engineering training to program for computers.

As a baseline, I used the ITU 2021 report  on computing skills among the general population. It contains data for 91 countries between 2016 and 2020. The results are provided as a spreadsheet . In this spreadsheet, I averaged the statistics for each country for each skill. Then, I computed the weighted average for each skill, weighting the contribution of each country to the final average depending on the number of respondents in the present survey. It is to be noticed that the USA, Canada and Australia are absent from the database, which may bias the results. The following is therefore the reference distribution that we should expect among darktable users :

The surveying logic is different there since it focuses on non-exclusive tasks to achieve rather than on a step-by-step levels scale, but we can try to convert it into the same logic :

  • 5.9 % of the population is able to program (Levels 4 and 5, programmers), and we shall assume they know how to do all the other tasks as well.
  • 51.2 % of the population is able to download and install software, meaning 48.8 % cannot. These will be our 0. Basic user with no admin skills,
  • 36.7 % of the population is able to connect/install devices, meaning 63.3 % can’t. Assuming people who know how to connect device should also know how to install software, we can deduce that these 36.7 % include programmers, power-users and basic users with admin skills. Therefore, the level 1. Basic user able to run update is composed of the 63.3% of people who can’t connect devices, minus the 48.8% of people who can’t install software, which leaves us with 14.5 %,
  • we will not be able to split basic users with admin skills from power users, which will be taken as a group by subtracting the ratio of programmers from the ratio of people able to connect devices, that is 30.8 %.
Side note

The figures here are telling us that what most open-source projects consider “normal” or “basic” (adding software repositories, editing config files, using command-line) is totally unrealistic for a general population where 49 % can’t even download and double-click a .exe package to install it.

We can conclude that, compared to the general population :

  • the share of programmers (either high or low level) among the respondents is 7.8 times higher,
  • the share of users with admin skills and power users is 1.5 times higher.
  • the share of people able to run updates on their computer is 1.8 times higher,

I want to redo the same density 2D heat map to see if the computer skills are linked to the studying background.

We see that 36 to 39 % of the respondents have both a programming level and a scientific background and the skill level repartition is heavily asymmetric : almost half the scientists have a low-level programmer level. Scientists account for 75 to 78 % of the programmers.

For non-scientists, the most represented skill group is the power users. Respondents with an art background tend to be biased toward the least skilled, although the largest group of respondents with an art background is the power users in both native and non-native groups. Respondents with a technical background tend to be biased toward the most skilled.

Again, the results are surprisingly close for both groups, despite slight deviations in age and education level (the non-native being in average 10 years younger and more educated).

Then, I want to assess if the computational level can be correlated to the education level.

These results are interesting since they are the first to show a clear difference between the native and the non-native groups : the native group shows a weak correlation between education level and computational skills, though most programmers actually have no more than a bachelor degree, while the non-native group shows no correlation but a diffuse distribution.

In both groups, the bulk of power users and programmers is at the engineer level (bachelor for native speakers, master’s for non-natives), and longer studies don’t lead to better computational skills, which means computational skills are linked to the studying background. If we recall that, in both groups, people with a PhD represent around 13 % of the respondents, we see here that 54% of them (7 % of the respondents) have a programmer level.

The heatmaps given above show where the global population of respondents is distributed on a 2D plane. Now, I want to examinate the probability of users to have certain computer skills within each class, that is how the skills are distributed in each class of studying background. The first 2 lines are the reference population taken from the ITU 2021 survey (see above), and the corrected reference assuming people who can’t install software can’t install darktable and are not relevant here.

First, let us recall here that the military and law fields have each 15 respondents in total, between both groups, which is a lot less significative than the other groups.

These graphs make it abundantly clear that programmers are over-represented in every field, while the least skilled users are under-represented (even discarding users who can’t install software). Even non-technical fields like arts and humanities show a ratio of low-level programmers alone greater than the ratio of all programmers in the corrected population, and a lower ratio of least skilled users than the general population.

It is an understatement to say that darktable filters out the least skilled users and attracts the most skilled. Not only does it fail to attract artists (8 to 16% of the total respondents for graphic and other arts), but the kind of artist it attracts is anyway more computer-literate than the average population.

#Operating system

darktable comes from the Linux ecosystem, and users running Linux don’t have the choice to use the most notorious photo editing applications. But it was made available on Mac OS since the beginning, and on Windows since 2017 or so. It is often said that the bias toward computer-literate users is due to the bias toward Linux users. I want he to verify this assertion.

I asked respondents on what OS they do run darktable, allowing more than one answer per respondent. 1317 answers were given.

The large majority of users indeed runs Linux, 68 % of the natives and 72 % of the non-natives, but a significant part uses Windows too : 42 % of the natives and 38 % of the non-natives.

Then I want to see if we can find correlations between computational skills and OS used.

In both group, we find 40 % of the respondents being both Linux users and programmers. That said, only 64 to 69 % of the low-level programmers actually run Linux. We also find very similar rates of basic users able to run updates on Linux (native : 4.9 %, non-native : 4.9 %) and Windows 8 and newer (native : 5.6 %, non-native : 4.6 %).

We find that Windows 8 users have twice as many less-skilled users (level 1.) than Linux, in both groups. But comparing to the general population corrected, Windows 8 still has half as many less-skilled users.

Linux users have around 56% of programmers in both groups. But Mac and Windows users show the same biasing trend toward programmers as Linux, that is at least 31 % of programmers among Windows 8 users and 37 % among Mac users, which is in average 3 times as much as in the general population corrected. Mac OS ranks between Linux and Windows, both for the proportion of programmers and the proportion of less-skilled users, which contradicts the common belief that Mac is mostly used by people who do not care about their computer.

All OS therefore show the same trend : 16 % to 56 % of programmers where we expect no more than 12 %, 7 % to 14 % of less-skilled users where we expect at least 28 % (considering the corrected general population). The difference between OS is only in the magnitude of the bias.

We notice here that Linux is the predominant OS for all fields and for both groups. Let us now redo the same exercise but with probabilities normalized within classes instead of globally. The reference of OS market share for the general population is taken from StatsCounter.com  for March 2020 and corrected for desktop OS only.

Again, let us recall that the military and law field have both 15 respondents in total, which is significantly fewer than other fields. For the native group, the market share of Linux is very consistent across all fields : 48 to 58 %. Non-technical fields still have between 46 and 52 % of Linux users in their ranks. Windows and Mac share the rest, with a surprisingly high market share of Mac OS among scientists (7 %, more than non-graphic artists).

The non-native groups shows a larger share of Linux users (up to 65 % for scientists), and also some users on Free BSD (and not necessarily where we expect them), but Windows still occupies at least 27 % of the market for every field. But, along with the higher education level, higher rate of scientists, this higher inclination for Linux may again point toward a biased sample selected by its ability to speak English as a second or third language.

Overall, we see here that the distribution of computing skills is very consistent across operating systems : heavily-biased toward programmers and power users. The distribution of operating systems across studying fields is even more consistent, with roughly half of each field using Linux and roughly a third using Windows. We cannot blame Linux for biasing results toward programmers, the Windows and Mac users show the same bias, although to a lesser extent.

This means that we can’t find significative differences in the type of population or in the computing skills linked to the type of OS used : darktable users are of the same kind no matter their OS or background. The characteristic they all share is their above-average computational skills. The archetypal user of darktable is a programmer or a power user, and the fact that there are more scientists than artists in the user base means those power users and programmers are statistically harder to find among the non-technical/non-scientist crowd.

This confirms that darktable simply deters computer-illiterate people, regardless of their OS or studying background.

#User’s relation to photography

#Photography education

The next key factor to understand darktable’s users is to learn how they learned photography. To the question “How did you learn photography ?”, the possible answers were :

  • All alone, with no external resources (self-taught)
  • Alone on the Internet (self-taught),
  • Alone with books (self-taught),
  • With informal training (workshops, vocational classes),
  • With formal training (school, university),
  • I did not learn photography.

Respondents could give more than one answer, and 1695 responses were recorded.

The vast majority of respondents are self-taught, and a small part (around 25 %) has got any kind of teacher. This may indicate that photography is seen as easy-enough to not require the help of a teacher or that the respondents prefer to work alone.

68 to 71 % of respondents have learned photography at least partly on the internet. 35 % of the native group has had only the internet as an education material. 22 % of the native group has had either internship, formal or informal training.

These results are concerning because the resources available on the internet are usually provided by amateurs and often lack historical and technical accuracy, but also because things like digital color management cannot be invented out of thin air and darktable is really not designed to handle these things automatically, in place of an educated user. There is an important gap between the amount of knowledge expected from the typical darktable user and the background real users actually have, if their primary source of education is unchecked/unedited internet information.

Then, I want to split these results within classes of background fields to see if some trend can be identified.

These results are concerning in their own way because, on one side, darktable has users more educated than the average (in unrelated fields), but on the other hand, in both groups (native and non-native), no more than 25 % of users have actual training (formal or informal) in photography, which means 75 % of users consider photography easy-enough to ditch classes or are simply not motivated enough to consider taking them.

Given how little darktable does for user-friendliness and how many parameters it gives to users to configure themselves (assuming they know what they are about), this is off to a bad start and can only trigger frustration. We simply cannot expect users to get all the required knowledge by themselves, given how much wrong information is available on the internet.

#Photographic endeavour

I asked users what photography was to them. More than one answer per respondent was allowed and 1433 answers were recorded.

We see that photography is a serious hobby for most of the user base. Let us see how this is distributed among backgrounds :

6.3 % of the non-native group and 9 % of the native group practice photography as a part-time job. 1.6 % of the non-native group and 3.7 % of the native group do it as a full-time job.

These figures suggest that professional photographers from non-English countries are less likely to have an sufficient English level as to take the survey.

Since respondents could give more than one answer, the ratio of people who answered “part-time job” or “full-time job” or both is 10.7 % among the natives, and 7.4 % among the non-natives.

Let us see now how the compare between groups of same background :

With no surprise, we find the most professional among the user with a graphic art background, then humanities and – surprisingly — business and administration, which might indicate late professional reconversions. The main endeavour for all groups remains a serious hobby.

Now I want to see if there is a pattern matching the photographic endeavour of photographers with how they learned photography.

The internet is the main source of education for all kinds of endeavour, except for the class of the native speakers whose full-time job is photography. The second source is books.

Only 30 % of the native full-time professional photographers have interned with a professional, and that figure is down to 8.5 % for part-time ones, which represents 1.6 % of the user base.

30 % of the native full-time professionals and 13 % of the part-time ones have had formal training, which represents 2 % of the user base.

40 % of the native full-time professionals and 17.4 % of the part-time ones have had informal training, which represents 2.6 % of the user base.

Overall, the part of users having had either internship, formal or informal training is 22.4 % on the global native group, and 39.4 % for the sub-group of native professional photographers.

All this means that :

  • 78 % of the general user base is entirely self-taught,
  • 82 % of the non-professional users are entirely self-taught,
  • 61 % of the professional users are entirely self-taught.

Again, given the lack of curation and the overall low quality of internet ressources regarding accuracy and concepts, this is concerning with regard to the technical level in darktable.

#Keywords

I asked respondents what words described their photography the best. More than answer could be given. 2634 responses were collected.

The proportion of “figuring it out” decreases almost linearly with the seriousness of the training for the non-native group, but the trend is less clear for the native group.

The proportion of “technical” decreases almost linearly with the seriousness of the training for the native group, but the trend is, again, not clear for the non-native group.

“Personal”, “artistic” and “creative” are consistently the most recurring keywords among all groups. “Commercial” and “professional” are a minority. This is consistent with the proportion of hobbyists.

#Conclusion of this part

This is bad.

A photo processing software should attract a distribution of users that matches the distribution of real-life photographers. Instead, we find :

Group of people% of the general population% of photographers% of darktable users
Women49.6 %60-70 %< 2%
Master/PhD14.4 %-34.9 %
Programmers5.9 %-> 44 %
Windows users79.4 %-< 42 %
Mac OS users18.8 %-< 12 %
Pro photographers--< 7 %

That is, mostly a population of professional programmers/scientists/engineers with above-average education and computer skills, self-taught in photography mostly on the internet.

Professional and trained photographers are very valuable to any imaging software because they are not only more demanding, but also more likely able to express the kind of image they are looking for. As such, they can provide qualified feedback on the gap between what they expect and what they get, or the challenges and slowdowns they faced while producing their result.

Amateurs, beginners and average photographers are more likely to stick with whatever they are able to get out of the software, with no defined editing goal, and the feedback they may provide will be limited to the easiness and enjoyment of the process. This kind of feedback is not actionnable and no design project can stem from it.

Amateurs also lack the proper color training to diagnose issues with their bare eyes, just like the general audience will not hear a slightly detuned string that will bother a musician. It’s no coincidence if darktable’s users rely so much on histograms and other scopes to (usually mis-) diagnose what is happening in the picture frame and making up problems that don’t really exist (while missing the actual ones). Seeing is not looking, reading a picture is an acquired skill.

The problem is actually quite simple : when an user reports a problem, you need to rule out the hypothesis of user error or misunderstanding. It is expected that people who struggle with photography at large will will struggle with the software as well, that says more about the user’s skills than about the software abilities. Having a reference sample of people experienced enough with photography gets you closer from actual data on the software itself.

The current community of darktable users, being made mostly of hobbyists with tech/science background, is of little use to help improving it as an artistic visual tool. And there is a real reconsideration that should be made regarding why the skilled and trained image retouchers choose other software. The fact they don’t need us is concerning because we certainly need them, to improve the software, showcase its abilities but also educate the rest of the user base.

The real problem, though, is that the small subset of users interacting with darktable development have to go through Github, which is a software forge deterring a lot of non-programmers. This, again, filters out a large part of the population, who doesn’t even understand what a software forge is and gets lost in the technicalities of “pull requests” and “issues”. As such, it biases even more the direction taken by the development because we basically drive away anybody non-technical, and solutions proposed by developers and validated by power-users or other developers constitute a strong confirmation bias regarding the abilities of the software.

The open-source/free/libre world is keen on its mantras, like data sovereignty, privacy, conservancy, freedom, etc. What the FLOSS advocates fail to see, though, is the solutions they offer and promote are largely inaccessible to the general population, and they need to be reminded of that every day. Open-source doesn’t discriminate on financial means, race, age or religion… It discriminates on computer-litteracy and apparently on gender.

I think the FLOSS communities and their advocates are largely deluding themselves regarding what they are and how exclusive they are, while pretending to be inclusive and open.

This work took a total of 55 hours and incurred 40 $ of expenses.

Next in this series : the results of the same survey ran again for 2022.