‘Master,’ ‘Slave’ and the Fight Over Offensive Terms in Computing (Kate Conger, New York Times, April 13, 2021)
The Linux Foundation | 06 May 2021
Nearly a year after the Internet Engineering Task Force took up a plan to replace words that could be considered racist, the debate is still raging.
Anyone who joined a video call during the pandemic probably has a global volunteer organization called the Internet Engineering Task Force to thank for making the technology work. The group, which helped create the technical foundations of the internet, designed the language that allows most video to run smoothly online. It made it possible for someone with a Gmail account to communicate with a friend who uses Yahoo, and for shoppers to safely enter their credit card information on e-commerce sites.
Now the organization is tackling an even thornier issue: getting rid of computer engineering terms that evoke racist history, like “master” and “slave” and “whitelist” and “blacklist.”
But what started as an earnest proposal has stalled as members of the task force have debated the history of slavery and the prevalence of racism in tech. Some companies and tech organizations have forged ahead anyway, raising the possibility that important technical terms will have different meanings to different people — a troubling proposition for an engineering world that needs broad agreement so technologies work together.
While the fight over terminology reflects the intractability of racial issues in society, it is also indicative of a peculiar organizational culture that relies on informal consensus to get things done.
The Internet Engineering Task Force eschews voting, and it often measures consensus by asking opposing factions of engineers to hum during meetings. The hums are then assessed by volume and ferocity. Vigorous humming, even from only a few people, could indicate strong disagreement, a sign that consensus has not yet been reached.
The I.E.T.F. has created rigorous standards for the internet and for itself. Until 2016, it required the documents in which its standards are published to be precisely 72 characters wide and 58 lines long, a format adapted from the era when programmers punched their code into paper cards and fed them into early IBM computers.
“We have big fights with each other, but our intent is always to reach consensus,” said Vint Cerf, one of the founders of the task force and a vice president at Google. “I think that the spirit of the I.E.T.F. still is that, if we’re going to do anything, let’s try to do it one way so that we can have a uniform expectation that things will function.”
The group is made up of about 7,000 volunteers from around the world. It has two full-time employees, an executive director and a spokesman, whose work is primarily funded by meeting dues and the registration fees of dot-org internet domains. It cannot force giants like Amazon or Apple to follow its guidance, but tech companies often choose to do so because the I.E.T.F. has created elegant solutions for engineering problems.
Its standards are hashed out during fierce debates on email lists and at in-person meetings. The group encourages participants to fight for what they believe is the best approach to a technical problem.
While shouting matches are not uncommon, the Internet Engineering Task Force is also a place where young technologists break into the industry. Attending meetings is a rite of passage, and engineers sometimes leverage their task force proposals into job offers from tech giants.
In June, against the backdrop of the Black Lives Matter protests, engineers at social media platforms, coding groups and international standards bodies re-examined their code and asked themselves: Was it racist? Some of their databases were called “masters” and were surrounded by “slaves,” which received information from the masters and answered queries on their behalf, preventing them from being overwhelmed. Others used “whitelists” and “blacklists” to filter content.
Mallory Knodel, the chief technology officer at the Center for Democracy and Technology, a policy organization, wrote a proposal suggesting that the task force use more neutral language. Invoking slavery was alienating potential I.E.T.F. volunteers, and the terms should be replaced with ones that more clearly described what the technology was doing, argued Ms. Knodel and the co-author of her proposal, Nielsten Oever, a postdoctoral researcher at the University of Amsterdam. “Blocklist” would explain what a blacklist does, and “primary” could replace “master,” they wrote.
On an email list, responses trickled in. Some were supportive. Others proposed revisions. And some were vehemently opposed. One respondent wrote that Ms. Knodel’s draft tried to construct a new “Ministry of Truth.”
Amid insults and accusations, many members announced that the battle had become too toxic and that they would abandon the discussion.
The pushback didn’t surprise Ms. Knodel, who had proposed similar changes in 2018 without gaining traction. The engineering community is “quite rigid and averse to these sorts of changes,” she said. “They are averse to conversations about community comportment, behavior — the human side of things.”
In July, the Internet Engineering Task Force’s steering group issued a rare statement about the draft from Ms. Knodel and Mr. ten Oever. “Exclusionary language is harmful,” it said.
A month later, two alternative proposals emerged. One came from Keith Moore, an I.E.T.F. contributor who initially backed Ms. Knodel’s draft before creating his own. His cautioned that fighting over language could bottleneck the group’s work and argued for minimizing disruption.
The other came from Bron Gondwana, the chief executive of the email company Fastmail, who said he had been motivated by the acid debate on the mailing list.
“I could see that there was no way we would reach a happy consensus,” he said. “So I tried to thread the needle.”
Mr. Gondwana suggested that the group should follow the tech industry’s example and avoid terms that would distract from technical advances.
Last month, the task force said it would create a new group to consider the three drafts and decide how to proceed, and members involved in the discussion appeared to favor Mr. Gondwana’s approach. Lars Eggert, the organization’s chair and the technical director for networking at the company NetApp, said he hoped guidance on terminology would be issued by the end of the year.
The rest of the industry isn’t waiting. The programming community that maintains MySQL, a type of database software, chose “source” and “replica” as replacements for “master” and “slave.” GitHub, the code repository owned by Microsoft, opted for “main” instead of “master.”
In July, Twitter also replaced a number of terms after Regynald Augustin, an engineer at the company, came across the word “slave” in Twitter’s code and advocated change.
But while the industry abandons objectionable terms, there is no consensus about which new words to use. Without guidance from the Internet Engineering Task Force or another standards body, engineers decide on their own. The World Wide Web Consortium, which sets guidelines for the web, updated its style guide last summer to “strongly encourage” members to avoid terms like “master” and “slave,” and the IEEE, an organization that sets standards for chips and other computing hardware, is weighing a similar change.
Other tech workers are trying to solve the problem by forming a clearinghouse for ideas about changing language.
That effort, the Inclusive Naming Initiative, aims to provide guidance to standards bodies and companies that want to change their terminology but don’t know where to begin.
The group got together while working on an open-source software project, Kubernetes, which like the I.E.T.F. accepts contributions from volunteers. Like many others in tech, it began the debate over terminology last summer.
“We saw this blank space,” said Priyanka Sharma, the general manager of the Cloud Native Computing Foundation, a nonprofit that manages Kubernetes. Ms. Sharma worked with several other Kubernetes contributors, including Stephen Augustus and Celeste Horgan, to create a rubric that suggests alternative words and guides people through the process of making changes without causing systems to break. Several major tech companies, including IBM and Cisco, have signed on to follow the guidance.
Although the Internet Engineering Task Force is moving more slowly, Mr. Eggert said it would eventually establish new guidelines. But the debate over the nature of racism — and whether the organization should weigh in on the matter — has continued on its mailing list.
In a subversion of an April Fools’ Day tradition within the group, several members submitted proposals mocking diversity efforts and the push to alter terminology in tech.
Two prank proposals were removed hours later because they were “racist and deeply disrespectful,” Mr. Eggert wrote in an email to task force participants, while a third remained up.
“We build consensus the hard way, so to speak, but in the end the consensus is usually stronger because people feel their opinions were reflected,” Mr. Eggert said. “I wish we could be faster, but on topics like this one that are controversial, it’s better to be slower.”
Kate Conger is a technology reporter in the San Francisco bureau, where she covers the gig economy and social media. @kateconger
Similar Articles
Browse Categories
Cloud Computing Compliance and Security Open Source Projects 2024 Linux How-To LF Research Open Source Ecosystem and Governance Blog Diversity & Inclusion Newsletter Data, AI, and Analytics linux blog Research Training and Certification Linux Cross Technology Cloud Native Computing Foundation cybersecurity software development Announcements Decentralized Technology Legal OpenSearch Sustainability and Green Initiatives cloud native generative AI lf events Finance and Business Technology Networking and Edge cncf industries Emerging Technology Health and Public Sector Interoperability Kubernetes Topic: Security Web Application & Development amazon web services aws community tools confidential computing challenges decentralized AI decentralized computing eBPF funding japan spotlight kernel license compliance openssf ospo research survey skills development state of open source tech talent