As part of the "Survey of human resource development in open source software engineering" currently being conducted on behalf of the Ministry of Economy, Trade and Industry, Mitsubishi Research Institute, Inc. conducted an online survey entitled "Free/Libre/Open Source Software Japanese Developers Online Survey." The objective of the survey was to determine the situation regarding open source software/free software (OSS/FS) engineers in Japan by gathering information directly from open source/free software developers themselves. The results will be used in personnel training in relevant technical fields, in planning policy for technology promotion and other areas.
This survey represents the Japanese version of the FLOSS survey* conducted in Europe in 2002 and the FLOSS-US survey** conducted in the United States during the first half of 2003. The FLOSS survey was conducted by the International Institute of Infonomics, University of Maastricht, Netherlands in 2002, under a grant from the IST Program of the European Commission. The FLOSS-US survey was conducted in 2003 by Stanford University's Stanford Institute for Economic Policy Research (SIEPR) under a grant from the National Science Foundation (NSF).
During a two-month period from September 1, 2003 through November 1, 2003, an online survey system was put up on the Mitsubishi Research Institute website (http://oss.mri.co.jp/floss-jp/). To prevent multiple responses, respondents were asked to first register an email address from the survey information page, and thereafter the actual survey URL was sent to that address. Survey forms were also distributed at open source related events. There were 55 questions in the survey, all of them multiple-choice. The estimated time required to complete the survey was 20-30 minutes. No particular restrictions were placed on who should respond to the survey, as long as the respondents considered themselves to be open source or free software developers who applied through several specific communities.
To enable global comparison with results obtained elsewhere, most of the questions were taken directly from the European FLOSS survey and the FLOSS-US survey. For this reason, the authors worked together with the leader of the FLOSS survey, Rishab Aiyer Ghosh, in developing the questions. The scope of the survey was as follows:
During the two-month survey period, 547 responses were received in all (487 online and 60 from people who filled out the form at special events). The number of responses was small in comparison to the 2,784 who completed the FLOSS survey and the 1,588 who completed the FLOSS-US survey. However, these earlier surveys targeted developers worldwide (in particular those in Europe and the United States). In contrast, the FLOSS-JP survey targeted only Japanese respondents, so considering this focus on a single country, it can be concluded that an adequate number of responses was received.
The section below will present a profile of the typical OSS/FS developer (gender, family composition, occupation, etc.).
First, a profile of OSS/FS developers (gender, family composition, occupation, etc.) will be presented.
Fig.1 (Q46) Gender
In general, many software developers are male. OSS/FS developers, too, are almost entirely male; only 2% of them are women. (Fig. 1)
These results were almost the same as those in the European and American surveys. Throughout the world, most OSS/FS developers are male.
Fig.2 (Q47) Family composition
It is said that most OSS/FS developers are single people who find it easy to secure time for development. Accordingly, we decided to ask the respondents about their family composition (single or married). As shown in Fig. 2, 61.5% of respondents were single. Based on the results of the 2000 national census, the average percentage of unmarried persons for the age distribution of survey respondents is 52.5%. The percentage of single respondents in this survey was 9 points higher, showing that many OSS/FS developers are unmarried. One factor may be that many OSS/FS developers also have high educational backgrounds, as will be discussed later.
The average age of unmarried respondents (without a specific partner) in the European and American surveys was 41.4% and 39.1%, respectively. The average age of unmarried respondents in Japan, 50.3%, differed somewhat from the results in the other surveys.
In Japan there is not usually a problem with asking the developer directly to state his or her age. In Western countries, however, this is avoided due to privacy considerations. In this survey as well, the authors decided to follow this precedent, and instead respondents were asked to state the number of years they had been involved in OSS/FS development and their age at the time they began the involvement. From this information, the current age of the respondent was estimated.
Fig.3 (Q4) Year respondent began involvement in OSS/FS development
Fig.4 Number of years of involvement
OSS/FS is by no means new. The free software movement dates back to around 1984, when Richard Stallman started to GNU Project. As shown in Figs. 3 and 4, the study found that, in Japan as well, about 10% of respondents began development around 1990. Still, active development is a recent phenomenon, with more than half (52.6%) of developers beginning their involvement in OSS/FS development in 2000 or later.
Fig.5 (Q5) Age of respondent at the time he/she began OSS/FS development
Figure 5 shows the distribution of the age of respondents when they began their involvement in OSS/FS development. At age 19 or later, involvement in development increases rapidly, with a peak at age 22 (7.1%). After that, the proportion decreases gradually, with only a few respondents beginning their involvement at or after age 40. These results suggest that many developers became interested in OSS/FS either during their college years when they had time to spare or after they had entered the work force and encountered OSS/FS in a work-related context.
Fig.6 Present age
Deriving the respondent's current age from the age at which he/she began development and the respondent's age at that time gives the distribution shown in Fig. 6. Respondents are active mainly in their mid-20s to mid-30s.
Table 1 shows the mean values for the age that the respondent began development and the respondent's current age, compared with the values for the surveys in Europe and the United States. The average developer began his or her involvement in OSS/FS development around the year 2000, when he or she was around 26, and the developer's current age is around 30. Compared to Europe and the United States, developers in Japan are three to four years older on average, and their period of involvement is shorter by about 1.5 years. From this it can be determined that OSS/FS became active in Japan about a year and a half later than in the West, and that most developers are not students but members of the work force.
Table 1 Mean/median values for year respondent began development, age at which respondent began development, and present age
| FLOSS-JP | FLOSS-EU | FLOSS-US | ||||
| Mean | Median | Mean | Median | Mean | Median | |
| Year respondent began development | 1998.4 | 2000 | 1996.7 | 1998 | 1996.6 | 1999 |
| Age at which respondent began development | 26.6 | 26 | 22.9 | 22 | - | 22 |
| Present age | 31.2 | 31 | 27.1 | 26 | - | 27 |
Fig.7 (Q48) Academic background
Figure 7 shows the academic background of respondents. The education level of respondents is quite high; approximately one third (33.6%) have completed graduate school (and have their master's degree or PhD). It can be inferred that many developers began development as a hobby in their undergraduate or graduate years when they had time to spare. On the other hand, the proportion of respondents with only a junior high school or high school education is also high (22.5%). Therefore, developers involved in OSS/FS development have wide-ranging academic backgrounds. This indicates that an academic background is not necessarily needed for OSS/FS development, and that anyone with the will to do so can become involved.
In Europe and the United States as well, the majority of developers had a high education level. This was particularly true of developers in the United States, in which 42.9% of survey respondents had a master's degree or higher.
Fig.8 (Q49) Occupation
Many developers are in software-related fields, with 41.1% of them describing themselves as software engineers and 10.0% as programmers (Fig. 8). This means that about half of the developers are involved in software development not only as a hobby but in a work capacity as well. In recent years, the number of people who develop OSS/FS in their job has been increasing, but as will be discussed later this still constitutes a minority. Accordingly, most people who engage in OSS/FS development do so outside of work. 14.4% are currently enrolled as students, with more of them majoring in subjects other than information-related subjects (8.0%) than those majoring in information-related subjects (6.5%). Again, this seems to indicate that one can become involved in OSS/FS development despite a lack of formal education in software development. 7.8% of respondents were affiliated with university or research organizations. Among these, there were some, although few in number, who released publicly as OSS/FS the software they had developed as part of their research activities.
The survey divided respondents according to information and non-information related fields. However, in statistical terms having a small number of respondents would result in greater variations, so for further analysis it was decided to organize respondents' occupations as shown in Table 2. The results organized in this manner are shown in Fig. 9.
As in the European and American surveys, many developers are software engineers. However, there is a great disparity with these surveys in terms of the number of students. The proportion of respondents who were students was 20.9% in Europe and 28.8% in the United States. These figures are 1.5 to 2 times the proportion in Japan. Moreover, in Europe, 3/4 of the students are in information-related subjects; this contrasts with the figure for Japan in particular. From this it can be inferred that there is a difference between Japan and the West in terms of information-related education. It should be noted that the American survey made no distinction between information-related and non-information related studies.
Fig.9 Respondent occupation (categorized)
Table 2 Categories for respondent occupation
| Software engineer | Software engineer |
| Student | Student (IT) |
| Student (other sectors) | |
| Programmer | Programmer |
| University / Research Institute | University / Research Institute (IT) |
| University / Research Institute (other sectors) | |
| Engineer (other sectors) | Engineer (other sectors) |
| Other | Consultant (IT) |
| Consultant (other sectors) | |
| Manager (IT) | |
| Manager (other sectors) | |
| Executive (IT) | |
| Executive (other sectors) | |
| Marketing / Product sales (IT) | |
| Marketing / Product sales (other sectors) | |
| Other (IT) | |
| Other (other sectors) |
Fig.10 Developer age distribution by occupation
Figure 10 shows the age distribution of the respondents in each occupational field. Naturally, the overwhelming majority of students are young. The age level is highest among engineers in non-information related fields. This is probably because respondents acquired software development expertise gradually as it became necessary in their jobs.
Fig.11 (Q50) Developer income
Fig.12 Developer income distribution by occupation
The income of developers varied considerably (Fig. 11). This was affected by the low annual income of the students who made up 14.4% of all developers. In actual fact, when income is viewed in terms of occupation (Fig. 12), it can be seen that 87.2% of all students had an annual income of no more than 1.2 million yen. Conversely, engineers in non-information related fields had a comparatively high income. This is probably because their age level is also high.
Table 3 (Q51) Nationality
| Japan | 99.8% |
| U.S. | 0.2% |
Table 4 (Q52) Current address
| Japan | 97.4% |
| U.S. | 1.3% |
| Germany | 0.4% |
| Other countries | 0.9% |
Table 3 and 4 show the nationality and current country of residence of the respondents. This survey targeted Japanese OSS/FS developers, and in fact almost all of the respondents listed Japan as both their country of origin and their current country of residence. The current country of residence also included responses such as Afghanistan and Algeria. As these countries were at the very top of the list of selections, it is very likely that these responses were entered by mistake.
Fig.13 (Q53) Country of origin/(Q54) prefecture of residence during student days/(Q55) current address (top: total/bottom: online survey only)
The remainder of this section will focus on changes in residence: area of origin -> area of residence during student days -> current area of residence. Figure 13 (top) shows the changes in respondent's area of residence. As survey forms were also collected at special events held in Tokyo and Osaka, it is possible that this increased the number of respondents in the Tokyo and Osaka areas. To avoid this problem, the lower graph shows only the tabulated results for the online survey.
There is a strong tendency for respondents to be centered in the Tokyo, Kanagawa, Saitama and Chiba areas, or the areas in the southern part of the Tokyo metropolitan area. There was a particularly rapid increase in the number of respondents who currently reside in Kanagawa Prefecture; this is probably the impact of the presence of many IT vendors that employ developers. The figures for Osaka Prefecture and Aichi Prefecture are almost unchanged, but a particularly noteworthy proportion of developers lived in Kyoto Prefecture during their student days. This is thought to be a sign of the active student club activities centered on Kyoto University and the like. There is also a great drop in the number of respondents who currently reside in Hokkaido, indicating that OSS/FS developers move away when they enter the work force.
Strictly speaking, open source software and free software are different. The free software movement is a kind of social movement, epitomized by the slogan "free all software" created around 1984 by Richard Stallman. It is known for the General Public License (GPL) and has a strong ideological color and imposes strict conditions (but on the other hand ensures freedom on an ongoing basis). Open source software, on the other hand, is a comparatively recent trend. In 1998, Linux and the like began to be used in the business world, and at that time the open source movement became one whose ideological aspects were de-emphasized and whose business advantages, such as "bazaar-like" joint development and free use of software, were emphasized. However, most developers also recognize the other group and cooperate with one another.
In this survey, respondents were asked which community they consider themselves to be a member of, and whether they thought there were differences between the two communities. Moreover, it was thought that this view would also be reflected in the licensing systems of the software they created, and so respondents were also asked what kind of licensing system they used.
Fig.14 (Q1) Member of open source community or free software community
First, respondents were asked whether or not they made a distinction between open source and free software and, if so, to which group they felt they belonged. Most (70.4%) respondents answered that they did make a distinction between open source and free software. In addition, a majority identified themselves with the open source rather than the free software community (Fig. 14); the figures were 43.7% for open source and 26.7% for free software. It was also learned that 29.6% were not concerned as to the group with which they were affiliated. In the European survey, 48.0% were in the free software community and 32.6% were in the open source community. In the American survey, the percentages for free software and open source were almost exactly equal (31.4% and 31.5%). Thus each of the three surveys found different responses for this topic.
Fig.15 (Q2) Differences between FS community and OS community
Fig.16 Differences in opinion according to affiliation
Next, respondents were asked whether the free software community and the open source community were the same or different. As Fig. 15 shows, almost an equal third of the respondents gave each of the three responses: "no difference" "different both in principle and in practice" and "the same in practice." However, most of the respondents who made no distinction between the communities were neutral, not allying themselves with either the open source community or the free software community, whereas most of the respondents who allied themselves with one of the two communities (81.6% open source and 85.6% free software) said there were differences between them (Fig. 16). The recognition of differences between the two communities was similar; slightly more respondents answered that the communities were different both in principle and in practice (47.7%) than those who answered that the communities were different in principle but the same in practice (45.9%). These results show that the respondents involved in OSS/FS development who made a distinction between the two communities did not do so out of a vague sense but with a clear understanding of the differences between free software and open source ideology.
Fig.17 (Q3) Favored licensing system
Fig.18 Differences in licensing system according to affiliation
Finally, respondents were asked which licensing system they would select if the code they created was not dependent on other code and they could select any licensing system (Fig. 17). Licensing systems tend to be considered troublesome or difficult, and only a small percentage (11.5%) of developers responded that they did not particularly care which licensing system they used. The most popular licensing system was GPL compatible (42.0%), followed by BSD style (30.2%). In terms of affiliation (Fig. 18), GPL compatible was the most popular licensing system for both the open source and the free software community, and more than half (56.8%) of the open source community in particular chose GPL compatible. Those who were indifferent to the distinction between the two communities responded with a variety of licensing systems, but there was a strong tendency for them to select the BSD style (37%). Respondents seemed to want to keep their distance from the GPL due to its strong ideological color.
Fig.19 (Q6) Average weekly OSS/FS development time
OSS/FS developers have an image of immersing themselves in development for long periods of time after work and on weekends. However, the amount of time that they actually spend on OSS/FS development is surprisingly short (Fig. 19). 61.7% of respondents said they spend 5 hours or less. In other words, in terms of time, many developers engage in OSS/FS development as one of their hobbies. Naturally at the peak of development they may engage in development for extended periods, but once the major development stage is over, routine maintenance may only require a brief period of time. As will be noted later, however, 90.8% of respondents said they use OSS/FS in their work or school activities, indicating that respondents are involved in OSS/FS development not merely as a hobby but as something that can help them in their work in some way.
On the other hand, 5.2% of respondents indicated that they spend 41 hours or more on OSS/FS development, and it is highly likely that these respondents are engaged in OSS/FS development as a business.
Fig.20 (Q7) Average weekly development time for proprietary software
53.1% of respondents develop proprietary software in addition to OSS/FS (Fig. 20). However, only 11.7% develop proprietary software on a full-time basis (40 hours a week or more); the majority (15.0%) spend 21-40 hours a week on this endeavor.
Fig.21 OSS/FS development time and proprietary software development time
There was no clear correlation between OSS/FS development time and proprietary software development time (Fig. 21). However, the respondents who spent two hours or less per week developing proprietary software also spent little time on OSS/FS development.
Fig.22 OSS/FS development time by occupation
Figure 22 shows the differences in time spent on OSS/FS development according to occupation. Programmers and persons affiliated with university or research organizations spent slightly more time on development than average. Conversely, the vast majority (55.1%) of engineers spent two hours a week or less on OSS/FS development.
Fig.23 Proprietary software development time by occupation
As could be expected, about 70% of software engineers and programmers, who presumably develop proprietary software as part of their job, develop proprietary software (Fig. 23). Nearly half (44.8%) of all engineers do so as well; these respondents presumably develop software for specialized purposes.
Fig.24 OSS/FS development time by family composition
OSS/FS development time was also examined in terms of family composition (Fig. 24). However, it does not appear that single persons without partners-who could be presumed to have an abundance of free time-developed for long periods of time. In fact, it was single developers with partners who spent the longest time in development. Single persons without partners and married persons spent about the same amount of time in development activities.
Fig.25 (Q8) OSS/FS development target fields
Networks and web services accounted for the largest percentages of development target fields; multimedia-related activities accounted for only a small percentage (Fig. 25). Most of the networks and web services consisted of small programs such as scripts and utilities. This is thought to be because such programs are easy to develop.
The order of target fields was exactly the same as the European survey. However, the figures for both network and web services were ten points or more higher in Japan than in Europe.
To determine the development languages and tools that developers used in OSS/FS development, respondents were asked to state their development platform, languages and tools used, and so on.
Fig.26 (Q9) Favored developing platform
Linux (RedHat, Debian, Vine, Turbo, Plamo, Gentoo, Mandrake, SuSE, Slackware and other Linux varieties) was the most frequently used platform (50.7%) (Fig. 26). Windows was next popular at 31.2%. Linux has a strong association with open source software, but in Japan many people develop on the Windows environment. This is particularly striking in view of the fact that Windows was listed by only 2.2% of respondents in the European survey.
Fig.27 (Q10) Platform used when respondent first began programming
The platforms that most of the respondents used when they first began programming were DOS (22.3%) and Windows (16.1%) (Fig. 27). However, 33.8% of respondents answered "Other." This category is thought to include such platforms as N88-BASIC, which was bundled with the NEC PC-9801 series, as well as pocket computers and so on.
Fig.28 (Q11) Platform used when respondent first began OSS/FS development
As in the case of the platform most often used, Linux and Windows were the most popular responses for the platform the respondent used when he/she first began OSS/FS development. However, the percentages were somewhat lower (Fig. 28). This is thought to be because the platform changed from DOS and Solaris to Windows and Linux.
Fig.29 (Q12) Development languages/tools with which respondent has experience
Fig.30 Programming languages with which respondent has experience
Major development languages such as C, C++ and Java were the most common responses. In addition, 47.8% of respondents had experience with SQL, indicating that many developers are engaged in system development using databases (Fig. 29/Fig. 30).
Fig.31 Scripting languages with which respondent has experience
JavaScript and PHP came in 3rd and 4th, respectively, in the area of scripting language (Fig. 31). It is thought that, since "web services" was one of the most popular development targets, developers learned JavaScript and PHP in the course of their development activities.
Fig.32 Development tools, etc. with which respondent has experience
In terms of development tools, 54.2% of respondents said they use CVS, while 34.7% used XML/SGML. This is an indication of the high literacy level of respondents (Fig. 32).
Fig.33 (Q13) Most commonly used desktop environment
Windows is the most commonly used desktop environment (43.7%). However, the figure for respondents operating under the XWindow environment (including X+Window Manager, GNOME and KDE) was 46%, which would seem to indicate at least that these respondents considered the OSS/FS desktop environment to be practical (Fig. 34).
Fig.34 (Q14) Favored editor
The most popular editor was Emacs (41.0%) (Fig. 34). It is thought that other editors include those that run under Windows (Hidemaru, etc.).
Involvement in the OSS/FS community was surveyed from the standpoint of not only time-related aspects but in terms of projects as well. The survey asked respondents to state the number of projects in which the respondent had been involved and was currently involved, the number of developers with which he or she had had contact, and the respondent's leadership experience.
Fig.35 (Q15) Number of projects in which respondent has been involved
respondents had not been involved in very many projects in the past (Fig. 35). 79.2% of respondents said that their past experience consisted of five projects or fewer, and 87.0% said they had been involved in no more than 10 projects. The average number was two to three projects. This seems to be an appropriate number since the mean for the number of years of experience was four to five years. Most respondents tended to have been involved in several projects rather than devoting themselves to a single project. Some respondents had been involved in more than 100 projects. It is assumed that these developers had been involved in Debian and FreeBSD/NetBSD package management.
Fig.36 (Q16) Number of projects in which respondent is currently involved
Respondents were not currently involved in many projects. The number of respondents answering that they were currently involved in one and two projects was 35.9% and 21.3%, respectively. 20.5% of respondents were not currently involved in any projects (Fig. 36). 4.6% of respondents were involved in six or more projects.
Fig.37 Relationship between current age and cumulative number of projects
A look at the number of projects in which the respondent has been involved in terms of age (Fig. 37) shows that respondents are actively involved at the age of 24-26. Only 2.5% of respondents replied that they had not been involved in any projects, whereas 15.0% had been involved in 10 projects or more. For the category of persons 36 years of age or over, the highest percentage of respondents in any age group (16.8%) indicated that they had been involved in 10 or more projects. However, the percentage of respondents who indicated that they had not been involved in any projects was also highest in any age group (19.6%). Thus there was a large gap between active developers and those who were not active.
Fig.38 Relationship between number of years of involvement and cumulative number of projects
A look at the relationship between number of years of involvement in OSS/FS development and the cumulative number of projects in which the respondent had been involved (Fig. 38) reveals that, as could be expected, the number of projects is larger for developers with a greater number of years' involvement. The exception is the period of 11 years' experience or more. For this category, the proportion of respondents without involvement in a single project is second highest (17.3%), second only to the value for developers with less than a year of experience.
Fig.39 Relationship between occupation and cumulative number of projects
There was no clear correlation between occupation and the cumulative number of projects (Fig. 39). The number of projects was slightly smaller for students, who had only been involved for a short period of time, and slightly larger for engineers and developers at universities and research organizations (in other than information-related fields).
Fig.40 (Q17) Number of projects in which respondent is currently involved as a leader
The percentage of persons with no leadership experience is slightly larger (56.6%), but the remaining respondents had had the experience of leading a project themselves (Fig. 40). In other words, a large number of developers have been involved in projects by heading the project themselves.
In the European survey, 35.2% of respondents had no leadership experience, a figure 21.4 points lower than the Japanese survey. This is thought to relate to the fact that the average period of involvement in development was 1.7 years longer and the median value was two years longer than in Japan.
Fig.41 Relationship between current age and leadership experience
A look at leadership experience in terms of age (Fig. 41) reveals that many developers aged 20 and younger and those 27-30 years of age have leadership experience. Developers aged 20 and younger are young and have not been involved in development for many years, so their leadership experience consists of two projects at most. However, the proportion of those with leadership experience is high in this age group (51.5%) and they actively head projects, so we can look forward to considerable development of OSS/FS from these developers.
Fig.42 Relationship between number of years' involvement and leadership experience
The greater the number of years' involvement on the part of the respondent, the more times the respondent had had leadership experience. However, the proportion of respondents with no leadership experience did not change very much (Fig. 42). It appears that persons who desire to lead a project themselves are able to do so immediately, whereas those who do not want to lead will continue to not do so. However, among developers in their eighth to tenth year of experience, one in three had no leadership experience, and this constituted the largest percentage. The cumulative number of projects was not particularly small, so it is thought that these developers participated in projects for which another developer conducted the main development work, in the capacity of translation into other languages and Japanese language localization.
Fig.43 Relationship between occupation and leadership experience
A look at leadership experience in terms of occupation (Fig. 43) reveals that developers at universities and research organizations have the most leadership experience. This shows that, since the software created by universities and research organizations is created for one's own research, inevitably there will be a greater number of projects for which the developers themselves serve as development leaders.
Fig.44 (Q18) Number of members of the community with which respondents have regular contact
To determine involvement in the OSS/FS communities, the survey attempted to determine how many people the respondents had contact with on a regular basis. Although some developers contacted 100 or more people, overall the number was comparatively low. 34.6% of respondents said they had contact with no one, and 33.3% said they had regular contact with one to five persons (Fig. 44). It is thought that the respondents who answered that they had contact with no one were engaged in creating and releasing small programs on an individual basis. Under such circumstances, it cannot be said that the merits of the open source style of development, which has been called the "bazaar model," are being utilized; there is a need for more active communication in order to develop larger projects.
Fig.45 Relationship between leadership experience and number of community members with which respondents have regular contact
In terms of leadership experience (Fig. 45), the more projects in which the respondent was involved, the more members with whom the respondent had contact.
Fig.46 Relationship between cumulative number of projects and number of community members with which respondents have regular contact
Viewed in terms of the number of projects the respondent has been involved in (Fig. 46), the number of persons with which respondents have contact increases as the number of projects increases. 40% of respondents who have been involved in six or more projects (40.4% for those involved in 6-10 projects and 50.0% for those involved in 11 or more projects) have regular contact with 11 or more people. These respondents can be said to be forming an active community.
Fig.47 (Q19) Main center of activities
There are no national borders in OSS/FS development, but naturally not all developers have contact with the global community. 62.3% of respondents said their main center of activity was the community in their own country. However, 37.7% had some involvement with the global community to a greater or lesser extent (Fig. 47). In the world of ordinary hobbies, it is rare to find a field in which 40% of the people are involved in global activities. It is likely that the opportunity to participate in global activities is one of the motives for participation in OSS/FS development.
Fig.48 (Q21) Ability to read, write and speak English
Fig.49 Relationship between English ability and communities in which respondents participate
A minimum level of English proficiency is needed to participate in the global community; however, Japanese are said to be lacking in English ability. In actuality, many developers (68.3%) feel their English writing ability is poor (Fig. 48). The community in which the respondents are involved differs according to English ability, and the degree of participation in the global community differs dramatically depending on whether or not the respondent has the ability to handle English composition (Fig. 49) (64% of those whose English is good enough to enable them to contribute to mailing lists are involved in the global community, as opposed to 29.5% of those who are lacking in English conversation and composition ability). To participate in the global community, a developer needs enough English language capability to be able to persuade others using logical arguments. However, since a certain number of developers participate in the global community even though their English language ability is poor, it appears that participation is possible if one has sufficient courage and perseverance.
Fig.50 (Q20) Major types of activity in which respondent was involved in OSS/FS development
Of the major types of activities in which respondents are involved as part of OSS/FS development, coding-related activities (developing main functions, fixing bugs, creating patches, testing) were cited by a majority of respondents. Document preparation/translation and support activities each garnered percentages of 10-15%, revealing that many developers contributed to activities other than coding (Fig. 50). This indicates that there are various ways for developers to be involved in OSS/FS even if they are unable to write code.
Fig.51 (Q22) Known OSS/FS community leaders
To determine the key players in OSS/FS development, a list of well-known OS developers was presented and respondents were asked to check the names of each person that they had heard of (Fig. 51). However, three names of fictitious persons were also included (Martin Hoffstede, Angelo Roulini and Sal Vallinger) to enable dishonest responses to be excluded.
Three of the names were particularly well-known: Linus Torvalds, the Linux developer; Richard Stallman, the creator of GNU, and Eric Raymond, the pioneer of the open source movement. Even so, the results were 12-20 points lower than the European survey; moreover, 11.3% of respondents answered that they did not know any of the names, indicating that Japanese are not that aware of the key players. In addition, Miguel de Icaza, leader of the GNOME project who came in third in the European survey with a score of 82%, had remarkably little name recognition in Japan at 17.9%. As described below, the other persons are also leaders in the OSS/FS movement.
Fig.52 (Q23) Uses for OSS/FS in work or school activities
In response to an investigation of whether respondents used OSS/FS not only as a hobby but in work or school as well, 90.8% answered that they used OSS/FS in their work or school activities (Fig. 52). With regard to the purpose of use, the overwhelming majority answered that they used OSS/FS as end users of applications (83.2%) and as program development tools (82.2%) (Fig. 53).
Fig.53 (Q24) Fields of OSS/FS use
Fig.54 (Q25) OSS/FS qualifications possessed by respondents
Fig.55 (Q26) Places where respondents learned OSS/FS development
Fig.56 Relationship between learning methods and whether or not respondent possessed qualifications
Only 7.9% of respondents had qualifications related to OSS/FS (Fig. 54). This was related to the place at which the respondents learned OSS/FS development. The combined percentage for developers who taught themselves or learned from the community came to 71.1% (Fig. 55). Of these, the percentage of those developers who possessed qualifications came to only 6.9% and 2.2%, respectively (Fig. 56). It is thought that most developers started their involvement with OSS/FS as a hobby, and therefore they saw no particular advantage to acquiring qualifications. On the other hand, a comparatively high percentage of developers who had learned as part of company work activities or who studied at information-related universities possessed qualifications. It is thought that this is because it was advantageous for those developers to acquire qualifications in terms of working conditions and gaining employment.
Fig.57 (Q27, Q28) Motivation for involvement in OSS/FS development
Five questions were asked of the respondents: what was the motivation for yourself/others to become involved in OSS/FS development; what are your/others' expectations for the OSS/FS community; and what is the purpose of the OSS/FS community. To each of these questions, a majority of respondents answered to acquire and share knowledge and skills (Fig. 57). The most important motivation for OSS/FS was to help improve one's skills. Many respondents also said that their motivation was the development of software itself: to turn ideas for new software into reality or to improve the OSS/FS of other developers. And the third largest proportion of respondents said that they wanted to achieve things they could not achieve with proprietary software, and that they thought proprietary software was bad. This seems to indicate that the ideology of the free software movement has become instilled in OSS/FS developers. Comparing the responses for the respondent's own views with what the respondent thought were the views of others, it is intriguing that the responses for others seemed to place more emphasis on the good or bad nature of proprietary software; respondents themselves did not think so but thought that others were surely concerned about such matters.
In the European and American surveys as well, the acquisition and sharing of skills was a major motivation. However, in the American survey, the motivation with the highest percentage of responses was the one in which respondents said they wanted to be able to freely modify software themselves.
Fig.58 Differences between motivation for involvement in OSS/FS development due to affiliation
The motivation for involvement differed depending on the affiliation of the developer (Fig. 58). The members of the free software community emphasized sharing knowledge and skills and tended to believe that it was bad for software to be proprietary. This would seem to indicate that many developers agreed with the spirit of the free software movement. The members of the open source community tended to want to modify OSS/FS made by other developers, and to gain income. This is in line with the reasons that the open source movement developed. Developers who did not consider themselves to be in either community wanted to learn skills and realize new ideas; few said they wanted to share knowledge, participate in OSS/FS or improve their opportunities for business. It would seem that these respondents were involved purely because it was their hobby.
Fig.59 (Q29) Beginning of involvement in OSS/FS
Approximately half of all respondents said that their involvement began with coding, indicating that their first encounter with OSS/FS was releasing a program of their own (36.5%) or sending a patch (16.2%). The other half said their involvement began with communication or documentation-related activities (Fig. 59). As in the responses for type of activity, these responses indicate that involvement in OSS/FS does not consist solely of writing code.
Fig.60
(Q30) Expectations with respect to other OSS/FS developers
(Q31) Expectations of other developers with respect to oneself
As might be expected, many developers cited acquiring and sharing knowledge as the response to the questions regarding one's expectations of other developers and other developers' expectations of oneself (Fig. 60). On the other hand, responses such as "resolving problems that cannot be resolved using proprietary software" and "achieving new ideas for software products" were listed as expectations not with regard to oneself but with regard to others, showing the same trend as in the responses for the respondent's motivation for involvement in development activities.
Fig.61 (Q32) Purpose for the existence of the OSS/FS community
Respondents cited a variety of purposes for the OSS/FS community. These can be summed up as creating better and better software, not for competition or desire for material possessions, but in a manner that allows everyone to enjoy the process and improve by learning from others in the spirit of friendly competition (Fig. 61).
Fig.62 (Q33, Q34) Balance (for oneself/others) between the benefits obtained from the OSS/FS community and the contributions made to the OSS/FS community
Regarding the balance between benefits and contributions on the part of oneself and others, many respondents said they felt that the benefits they received exceeded the contributions they made to others (Fig. 62). The responses seemed to indicate that a good environment had been born in which people helped one another out through the community, in which most of the community members derived benefits and, moreover, became involved in the community.
A signature in the source code is proof that the work is one's own creation. Accordingly, developers were asked how they regarded signatures.
Fig.63 (Q35) Signature in source code
Fig.64 Relationship between age and attitude toward signature
Fig.65 Relationship between occupation and attitude toward signature
In response to a question regarding whether the respondent included his or her own signature in the source code, 92% said they did. However, only 36% said they included their real name (Fig. 63). In terms of age (Fig. 64), there was a strong tendency to use handles among respondents 23 or younger, and particularly 20 or younger (probably because many of these are students). And, in fact, when viewed in terms of occupation (Fig. 65), students do tend to like the use of handles. Moreover, possibly because they enjoy their work, programmers do not seem to be very concerned about signatures. On the other hand, awareness is high among engineers (in non-information related fields). This would seem to be because, due to the nature of their work, they are always aware of patent and other rights.
Fig.66 (Q36) Source of income (if any) related to OSS/FS and means of deriving income
Fig.67 OSS/FS related income in terms of occupation
Only 26.8% of developers derived income directly from OSS/FS activities (Fig. 66). More than 70% were involved in OSS/FS development purely as a hobby. Viewed in terms of occupation (Fig. 67), a comparatively high proportion of software engineers and programmers derived remuneration, in the form of compensation for software development.
Among the respondents who received remuneration, a slightly higher percentage received indirect, non-monetary remuneration (Fig. 68). 47.4% of respondents received no remuneration at all, whether direct or indirect. In the European survey, 53.7% of respondents received monetary assistance; in the American survey, this figure was 43.2%. This trend is expected to increase in Japan as well. Noteworthy with regard to indirect remuneration is that 9.6% of the respondents said their OSS/FS development experience helped them to gain employment or change jobs. This is because it indicates that OSS/FS development is starting to be recognized as experience for software developers. If developers are able to receive income from OSS/FS development itself in the future, this will be a solid resume credit when they seek to gain employment or change jobs.
Fig.68 (Q37) Indirect remuneration related to OSS/FS activities
Fig.69 Indirect remuneration related to OSS/FS activities in terms of occupation
Viewed in terms of occupation (Fig. 69), as in the case of the OSS/FS related sources of income, the proportion of students and engineers receiving indirect remuneration from OSS/FS activities is low. The highest proportion was found among developers at universities and research organizations. However, for all occupations, the largest number of respondents said they received remuneration that was not directly related to work.
Fig.70 (Q38 - 41) Number of times respondent received assistance
Fig.71 (Q38 - 41) Number of times respondent received assistance (for respondents who received assistance at least once)
The trend toward receiving assistance for OSS/FS activities from outside sources has been increasing in recent years (Fig. 70/Fig. 71). Prior to six years ago, only 2% of developers received external assistance. Three to five years ago, this had risen to approximately 12%, and within the past two years 20% of developers had come to receive assistance from outside sources. In all, 26.1% of respondents had received assistance in the past. The source of the assistance was most commonly government or other public funds, followed by hardware vendors, SIers, individual contributions, distributors, and schools and universities (Fig. 72). It can be said that OSS/FS developers have a wide variety of sources of assistance.
Government and public funds, the largest source of assistance in Japan, was 5th in the American survey. Private sector assistance appears to be less widely available in Japan than in the West.
Fig.72 (Q42) Source of assistance during the past five years
Fig.73 (Q45) Income derived from OSS/FS activities during the previous fiscal year
Most of the developers who currently derive income directly from OSS/FS activities derive 1.2 million yen or less (Fig. 73/12.3%). For most, OSS/FS development constitutes one portion of their work, and they do this concurrently with their other work tasks. Less than 10% of developers are able to live off what they earn from OSS/FS development alone. This shows how difficult it is to make a living through OSS/FS development alone. However, although few in number, some developers have an income of 6 million yen or more, so it can be said that in some companies OSS/FS is recognized as a bona-fide work task.
Fig.74 Affiliation in terms of OSS/FS related income
Fig.75 Licensing system used in terms of OSS/FS related income
The remainder of this section will take a detailed look at the views and affiliation of people who derive income from OSS/FS related activities. As there can be expected to be differences between developers who derive only pocket money from their OSS/FS activities and those for which OSS/FS represents a major source of income, the respondents were divided into three categories: those with no income, those with an annual income of less than 1.2 million yen from OSS/FS development, and those with an annual income of 1.2 million yen or more from OSS/FS development.
As Fig. 74 shows, many of the developers who derive no income from OSS/FS activities are not concerned about community affiliation, whereas most of the developers who derive income from OSS/FS are in the open source community. In terms of licensing systems, GPL compatible is popular and is used by more than half of the respondents (51.5% for those with an income of less than 1.2 million yen; 54.0% for those with an income of 1.2 million yen or more) (Fig. 75).
Fig.76 Motivation for involvement in OSS/FS development in terms of OSS/FS related income
With the exception of acquiring and sharing knowledge and skills, the motivation for involvement in OSS/FS activities differs according to income (Fig. 76). It is thought that developers who earn 1.2 million yen or more make ends meet through their OSS/FS related income, so motivations related to making a living can be expected to be a major factor. 40.0% of respondents said their motivation was to derive income, while 22.0% said it was to increase the opportunities for work. It can be said that the reason that they derive income from OSS/FS rather than proprietary software is that they feel it is bad for software to be proprietary (30.0%) and because they want to resolve problems that cannot be resolved using proprietary software (34.0%).
Developers who earn less than 1.2 million yen have little desire to develop new software themselves, but they do want to reduce their dissatisfaction with existing software. This may be because the percentage of respondents who said that their motivation is to turn new ideas for software products into reality (13.6%) is lower than average, whereas conversely the percentages of respondents who said that they want to improve the OSS/FS of other developers (30.3%), and those who feel there are limits to the capabilities of software companies (16.7%), are higher than average. The percentages of these developers who want to obtain income and work are 10.6% and 15.6%, respectively, values that are in the middle between those with no income and those who earn 1.2 million yen or more.
Respondents who derive no OSS/FS related income place particular emphasis on acquiring knowledge and skills (62.9%). They place less emphasis on acquiring income and work (3.8% and 10.5%, respectively), and on ideology with regard to software (proprietary software is bad: 11.4%), as compared to developers who derive income. They want to come in contact with new and unusual items, wanting to disseminate software with little market value (11.2%) and wanting to participate in a new kind of entity for joint development.
Fig.77 Location of OSS/FS activities in terms of OSS/FS related income
As income increases, the respondent's location for activities tends to become more global (Fig. 77). Only one third (33.3%) of those making zero yen or less are involved with the global community, but 58% of those making 1.2 million yen or more are involved in the global community.
Fig.78 Attitude toward signatures in terms of OSS/FS related income
There is a difference in attitude regarding signatures depending on whether or not the respondent has OSS/FS related income. The percentage of respondents who sign their real names is 29.8% for those with no income and 54.5% (less than 1.2 million yen) or 50.0% (1.2 million yen or more) for those who derive income from OSS/FS activities (Fig. 78). It seems likely that developers need to publicize their work in order to derive income, and for this reason they must include their signature.
Fig.79 Number of people with whom respondent is in contact, in terms of OSS/FS related income
The number of people contacted by the respondent tends to increase as the income of the respondent increases (Fig. 79). This is likely to be because it is difficult to derive income unless the project is of a certain scale, and as a result the number of project members increases.
To make an overall evaluation of the above survey results, it can be said that the developers who derive income from OSS/FS related activities are those who play a central role in the OSS/FS world. These developers are involved in the global community that includes both Japan and overseas countries and are in contact with many other people. Moreover, they think proprietary software is bad and favor GPL compatible licenses.
Fig.80 (Q43) Recognition of OSS/FS development (current)
Fig.81 (Q44) Recognition of OSS/FS development (desired)
Accordingly, the authors decided to try to determine whether companies (schools) were aware of the involvement of respondents in OSS/FS development. The results showed that 7.2% of respondents worked for OSS/FS development and received compensation (Fig. 80). However, most (40.8%) companies /schools did not know of the involvement of developers in OSS/FS activities, and the developers wanted schools/companies to recognize their involvement (29.4%) and wanted to receive compensation (24.2%) (Fig. 81). Developers actually wanted to be involved in OSS/FS as a work activity.
In the American survey, 59.2% of respondents said their company was aware of their involvement, while 22.6% said the company was unaware of their involvement, indicating that the degree of awareness on the part of companies was quite high. In addition, 27.4% said their company had assigned them to conduct OSS/FS as a work task, indicating the degree to which Japanese companies are behind in terms of recognition of OSS/FS development.
This could be called a gap in awareness of OSS/FS between companies and developers. There is a need for efforts on the part of companies to understand the advantages and disadvantages of OSS/FS development, and for efforts on the part of developers to make their endeavors known. However, individual developers have little power, so the community as a whole must act to influence public opinion. It is also important for national and local governments and the like that have shown a comparative degree of understanding with regard to OSS/FS to work to increase understanding.
The FLOSS-JP Free/Libre and Open Source Software Developers Online Survey, ran for a period of approximately two months from September 1 through November 1, 2003, and the results given in the preceding pages were obtained. This section sums up the findings to give an overall picture of OSS/FS developers in Japan.
Almost all developers (98%) are male, with the active age range being primarily mid-20s to mid-30s. The average developer is currently 31.2 years old and started developing at the age of 26.6, and has therefore been involved in OSS/FS development for 4.6 years. That makes Japanese developers three to four years older than those in the West, and their career is 1.5 years shorter. More than half (52.6%) of all developers began their development work in 2000 or later, but 10% started around 1990. This indicate that OSS/FS development in Japan is has quite a long history. Most developers are highly educated; about one in three has completed graduate school. However, 22.5% have only a junior high school or high school education, indicating that anyone can become involved in OSS/FS development given the will to do so. In terms of occupation, half of all developers are in software-related occupations. Fewer (14.5%) are students as compared to Europe (20.9%) and the United States (28.8%), and whereas in the West 3/4 of student developers are on courses related to information technology, these account for less than half of the total in Japan. In terms of current residence, developers are concentrated in the Tokyo area (southern Kanto) (51.6%).
In terms of their opinion regarding to free software and open source software, two-thirds of all respondents make a clear distinction between free software and open source software, and open source software is the more popular. In the European survey, free software was the more popular, while in the American survey the percentages were about equal, so each survey obtained different results on this question. With regard to licensing systems, most developers preferred the General Public License (GPL).
Looking at the time actually spent on OSS/FS activities, many developers treated their work like a hobby, spending five hours or less per week on it. Only a minority spent long periods of time on development. The most common development platform was Linux, but 30% developed under Windows - a clear difference with surveys in the West, in which only a few percent did so. Developers had on average only been involved in a few projects and had only led a few; however, just under half of the respondents claimed project leadership experience, so it was concluded that many had initiated projects themselves. For this reason, developers are not in close contact with many other members of the community, and so even though their source code are open, it is thought that their activities did not up to "bazaar-like" development.
Since most Japanese are unable to express themselves well in English, it was thought that they would be limited to participation in the domestic community, but in actuality nearly 40% of developers were involved with the global developer community. Many were involved with other developers outside Japan despite a lack of English composition ability, so overseas involvement appears possible given sufficient effort on the part of the developer.
Most developers said they were involved in OSS/FS development in order to improve their knowledge and skills, and in this they appear to be reaping the benefits desired. Though "skills" might be taken to mean programming skills, it was clear that many developers also contribute in areas other than programming, such as the preparation of documentation, support provision, and the like. About half listed software and patch releases as the beginning of their involvement in OSS/FS development, while approximately another half listed email contact with other software creators, bug reporting, document translation, and other activities as their initiation. An overwhelming majority of developers gained their expertise through self-study (62.5%), with 8.6% listing the developer community and 14.1% listing work experience as the source of their knowledge. Combining these three figures, it is revealed that 85% of developers have received no formal OSS/FS education such as seminars and university classes. Further, few have qualifications as OSS/FS developers.
Concern for signatures is low on the part of younger developers (students) and programmers, and education regarding copyright matters appears to be inadequate.
Few developers receive monetary assistance relating to OSS/FS development work (26.8%), fewer than in Europe (53.7%) and the United States (43.2%), but the number is gradually increasing. In addition, more developers are able to use their skills to gain employment or change jobs. However, the current situation is that there is a wide gap in recognition between developers and companies.
Based on the results of the survey, the authors finish this report by making the following recommendations for the encouragement of OSS/FS development in Japan.
Japanese OSS/FS developers enter the field when they are three to four years older than developers in the West, and the percentage of students involved is lower. This is thought to be because students at universities and trade schools have few opportunities to become involved with OSS/FS. This is also indicated by the fact that 30% develop in the Windows environment, as opposed to only a few percent of developers in the West. In particular, the fact that students of information technology represent a smaller proportion than others (whereas in the West 3/4 are studying information technology) clearly indicates a problem with university-based education in information technology subjects in Japan. Universities in the West teach programming through OSS/FS, and Japan should also incorporate such instruction into its curricula at the earliest possible opportunity. The curricula should include the basics of OSS/FS, such as copyright and licensing issues, as well as communications skills.
Approximately 60% of developers learned their OSS/FS skills on their own. While it is true that for many developers OSS/FS development is an extension of a hobby, a means of systematic learning must be provided so as to promote OSS/FS development in the future. Many developers cite improving their expertise as an objective for OSS/FS development, and for this reason, also there is a need for better seminars and texts for use not only in college curricula but also by members of the public. To this end, it is essential to establish standards for OSS/FS skills that provide a formalized structure for learning.
Whereas around half of developers in the West derive income from OSS/FS development, in Japan only 1/4 do so. In most cases, the developer's company or school does not even know that the person is involved in OSS/FS development. On the other hand, it seems that developers want their involvement to be recognized and want their company or school to approve compensation for their activities if appropriate. If the general public had a greater awareness of the contribution made by OSS/FS, OSS/FS developers would undoubtedly find their situation improving. For this reason, developer communities and national and local governments should work to increase awareness of OSS/FS development among the general public.
Support for OSS/FS development is increasing each year. However, in the current situation, in which business models for OSS/FS development have not yet been firmly established, the level of support is not adequate. The basic needed in this area is for an awareness on the part of the administration that promoting OSS/FS development will help drive the IT industry in Japan and increase the competitiveness of Japanese industry. Small-scale project startups and bringing prominent OSS/FS development into practical application are thought to be particularly important. Some kind of assistance should be provided over a period of years to enable ongoing improvements to software, while assistance is also required for document preparation, training of support engineers, and other activities not directly related to software programming.
It is the hope of the authors that the findings of this survey will help promote OSS/FS development and improve the circumstances of OSS/FS developers.
This survey would not have been possible without the cooperation of OSS/FS developers. The authors would like to offer their thanks to those developers who submitted responses and helped publicize the survey. Thanks also to SourceForge.jp and Slashdot.jp and others for publicity, to VA Linux Systems Japan for assistance in implementing the survey at the Linux Kernel Conference, and to the Japan Linux Association and the Kansai Open Source + Freeware 2003 Executive Committee for assistance in implementing the survey at the Linux Conference 2003/Kansai Open Source + Freeware 2003 Conference. Finally, the authors would like to thank Rishab Aiyer Ghosh, leader of the FLOSS survey, for kindly consenting to the use of the questions from that survey.
©2004 Mitsubishi Research Institute, Inc. |
Webmaster