How long these terms have been in use
“Master/slave” showed up in technical documentation as early as the 1960s, borrowed uncritically from electrical engineering, where it described a straightforward relationship between a controlling device and a controlled one. Nobody who wrote it down was making a political statement — it was just the word already in use for that kind of relationship. That’s true of most of the terms on this list. They weren’t chosen to offend anyone; they were inherited, unexamined, from whatever field technical writing borrowed its vocabulary from first.
What’s changed isn’t that anyone rediscovered malice in these words. What’s changed is that enough people pointed out the discomfort, enough maintainers agreed the discomfort was reasonable, and enough of the ecosystem followed along that the old term stopped being the default anyone reaches for first. That’s the actual mechanism behind every entry in this list: not a decree, a slow accumulation of independent decisions that eventually looked like consensus.
Which organizations changed which terms
Each of these replacements was adopted by a named standards body or major project, on a date you can check.
Allowlist / denylist. Chromium, the Internet Engineering Task Force, and a large share of the security industry moved away from “whitelist/blacklist” over 2018–2020. The IETF’s own drafts on terminology, along with security teams at companies including Google and Twitter, cite the same reasoning: the metaphor ties “white” to permitted and “black” to forbidden, which is worth avoiding when a precise, equally short alternative exists regardless of how any individual writer meant it.
Primary/replica. Python’s core development team renamed “master/slave” in its multiprocessing documentation in 2018. Django, Redis, and MySQL made comparable changes to their own replication terminology around the same period. The technical meaning didn’t change at all — a primary still accepts writes and a replica still mirrors it — only the label did.
Chair, spokesperson, firefighter, and the rest of the “-man” job titles. This one predates the tech industry entirely. US federal job classifications dropped gendered titles for exactly this reason starting in the 1970s, and most professional style guides — AP, Chicago, and the Microsoft Writing Style Guide among them — have carried the gender-neutral versions as their default recommendation for decades. If your documentation still says “chairman” or “policeman,” it’s not using dated tech jargon; it’s using dated general English.
Singular “they.” This is the oldest change on the list by a wide margin — grammarians have documented singular “they” in English since at least the 14th century, including in Shakespeare and Jane Austen. What’s recent is style-guide endorsement of it as the default generic pronoun over “he,” “he/she,” or “s/he”: the American Psychological Association added it to APA style in 2019, and the Chicago Manual of Style and AP Stylebook both followed within a few years.
Adoption is the criterion, not preference
The reason this article names sources instead of just asserting “these terms are better” is that “better” is doing too much work on its own. A checker — or a person — that flags language purely on a feeling of what sounds more respectful is making a values judgment dressed up as a fact, and it will get the edge cases wrong constantly. Citing who actually adopted a term and why turns an opinion into a piece of information the reader can independently check, push back on, or decide doesn’t apply to their situation.
That standard cuts both ways. It’s also the reason some proposed replacements — “black box testing” to “opaque box testing” is a common example — don’t appear on this site’s checker at all. The argument for the change exists and has some backers, but it hasn’t reached the same breadth of adoption as allowlist/denylist or primary/replica. A tool that treats a fringe proposal and an industry-wide shift as equally settled teaches its users to stop trusting it on either one.
The cases a wordlist genuinely can’t resolve
Some of the terms above have a second, unrelated meaning that has nothing to do with the reason for the change. “Master’s degree” and “master branch” both use “master” in a sense with no connection to the master/slave pairing — which is exactly why the checker never flags bare “master” at all, only the compound “master/slave” or “master-slave.” “Guys” addressing a group that genuinely is all men isn’t a mistake; a checker just can’t tell that from the sentence alone, so it says so directly instead of pretending otherwise.
This is the same discipline the style guide linter applies to word choice generally: cite the source, name the exception, and stay quiet where the tool genuinely doesn’t know enough to be useful.
Check your own writing
Run a document through the inclusive language checker to see technical terms, gendered defaults, ableist idioms, and loaded framing flagged with the specific alternative and the reasoning behind it — every finding a suggestion, none of them a verdict. If gendered job titles and pronouns are the more relevant half for your document, see writing without gendered defaults for that piece in more depth.