Microsoft Helping Out In Making The Linux Kernel Language More Inclusive

[pong]

It's weird how these "problematic" words existed for the entirety of computer science as a field completely free of prejudicial connotation until one day four years ago they suddenly did.

Computer science took off conceptually in the 1930s-1950s when electronics / mechanics / IT started to be used for data processing, controls, et. al. If you think that the 1930s-1950s and on were free of prejudicial cultural biases which were very solidly embodied in the big government / corporate institutions doing IT / CS / telecom. et. al. you're way off base.

If your point is "we've always done it that way, why change now", well, by that standard there'd be no positive evolution of anything else and "OS" would mean CP/M still and not LINUX/UNIX/...

[pong]

Let me quote the commit for you.



We're engineers. We're supposed to be capable beings. If I cannot remember that the new "client" is what we used to call "slave" a few years ago, I'm not sure I can still call myself an engineer.

Hear hear.

Of course arguably at some point we (collectively as engineers) should have some better ways for interface definitions and requirements driven design or similar so that they could just update the "i2c schema to v7" and all the code that "includes" that / derives from it would just either automatically adapt upon re build / regeneration or the discrepancies would be flagged for corresponding semi-automated refactoring.

Some company changes their name and there's like a million manual edits to all source code and documents and web pages to change it, it's kind of ridiculous.

And if it was actually a functionally relevant thing like say some enumerated protocol identifier like I2C_REVISION_COMPLIANCE that changes from 6 to 7 but since nothing shares the header file there's 100k manual code updates to re-express that from some specification PDF to RUST, C, C++, Javascript, ....

The people that just write an IT / protocol / interface specification for ISO / whatever and publish a PDF with 1200 barely human readable "requirements", "terms", "glossary", ... and then have no way to actually express the facts in machine readable / usable format seem misguided. Make the specifications for IT things that have to happen "on a computer" in a precisely formed computer usable grammar / language / meta-language.

[GrumpyLinuxUser]

I wonder how Epstein's pals referred to the children on his private island...

[mos87]

No wonder it's the big money behind this neo-commie madness.

[ssokolow]

One could do better or at least as well without any difficulty.

"peripheral" is usually one role which bears a connotation to being an "accessory" to some system which actively makes use of its capabilities, so although one could have variously "active" / autonomous peripherals, generally it's clear that in the basic case they're devices which respond to some API / interface and which perform functions accordingly.

Or "device", "function", "unit", "accessory", "sensor", "device", ....

Device driver -> device which performs the function.

Or "client" may not be so bad.

"Peripherals" or "accessories" may or may not support DMA. (The thing that made "bus mastering" expansion cards so expensive and desirable.) "Slaves" clearly do not.

"Active" is already overloaded with too many other meanings. "Device" and "unit" include the host PC. "Function" refers to an action or a piece of code, not a physical object. "Sensor" is far too orthogonal to the underlying implementation details (eg. You can use an ESP8266 or ESP32 to build a "sensor" that's a a standalone WiFi device and blindly screams measurements into the void as UDP packets without caring if anyone is listening.)

"Device driver" is so obviously already taken that I'm wondering if you're trolling.

"Client" implies a client-server or P2P topology where the "client" is either a standalone computing device (eg. a PC) or the client software running on such a device. (If the counterpart to a "server" isn't not capable of standing on its own, then it's a "terminal".)

Fundamentally, the thing that makes people uncomfortable about the term is exactly what it's meant to express. Slave devices are called "slaves" because they have no agency and only act on orders from the master device. Any new word that captures that is likely to just spin the euphemism treadmill and be considered problematic later.

(eg. a "controller/worker" topology doesn't preclude things like workers that engage in load balancing via work-stealing queueing and not all "slaves" make sense as "workers".)

[lyamc]

This forum discussion is evidence to me that making the change actually hurts more than it helps. If it were for technical or ease-of-use reasons I doubt many people would care. It's like a refactor that maintains the complexity and the same performance, while breaking some backwards compatibility. Why would anyone support that?

The problem with most of these changes is that they:

1. Add technical debt. A lot of documentation and peripheral scripts/apps have to be updated as well.
2. Redirect resources from other, useful things
3. Are actually more confusing to users (primary password)
4. Don't even accomplish their self-stated goal. Whenever someone does this, the last thing I think is "boy I sure would like to contribute to this project". The only thing it encourages is more people who make useless changes to pat themselves on the back. Updating documentation is useful, but updating documentation to change two words in order to encourage others to update documentation is just retarded.
5. Cedes authority to these ideologically captured people. Don't like someone? Accuse them of not being diverse enough until they're either kicked out or leave.

I'm not going to stop wearing pants just because Hitler wore pants.

[niner]

I'm an application developer. I'm speaking generally, on behalf of overworked developers who have seen this song-and-dance before and know it won't accomplish anything except wasting our unpaid time.

So this particular change does not affect you at all. The actual kernel maintainers accepted the change. Stands to reason that they are ok with it. Could even be that they welcomed it. Who are you to pretend to speak for them and disagree?

[Spacefish]

Now it´s even worse to understand i2c
MISO (Master In Slave Out) and MOSI (Master Out Slave In) made it clear compared to TX / RX..
Now we have Client / Server? But on the chip / chip datasheet the pins are still called MISO and MOSI
++confusion

[anda_skoa]

If it were for technical or ease-of-use reasons

Updating the names in the code to match the names in the related hardware specification documents, as stated in the submission email, would qualify for "technical or ease-of-use reasons", no?

[lyamc]

Updating the names in the code to match the names in the related hardware specification documents, as stated in the submission email, would qualify for "technical or ease-of-use reasons", no?

It depends. Did the hardware specification documentation change or was it always like this?

If it didn't change then it's perfectly reasonable. If it did change, then did it change for technical reasons or did it change because of some woke ideologues?