Sounds like your experience is primarily with data stores then?

https://en.wikipedia.org/wiki/Master/slave_(technology)

It's used correctly in several of the cases listed here. Also, Jenkins uses (or used?) the term correctly.

replica doesn't imply a power relationship. Master controls the slave , which is accurate

- Master database tells the slaves what to write

- Master clock tells slave the time

- Master HDD controls the computer , so indirectly controls secondary -- though in this case slave HDD is more auxiliary rather than subordinate (this terminology isnt used anymore anyway)

In any case, master/slave terminology is historical and generally unproblematic, except apparently in the US, but so many words are problematic for someone somewhere that it's not wise to go down that road

Not accurate in databases (where master only means authoritative source), not accurate in hard drives (master only denotes precedence), not accurate in hardware clocks (used to determine sync hierarchy - see NTP for a much better convention). "Master" only makes sense in a system where a component is delegating tasks and controlling one or more "slaves", and even then something like "controller" is clearer.

- Master database tells the slaves what to write

But that's only an ad-hoc state of affairs; the only thing telling a database (in postgres, at least) that it's a replica is that it has a couple of settings telling it what to replicate from, and if you remove those and restart (or otherwise trigger a failover), the replica becomes the new master.

In slavery, the only correct thing for a master to do is to abdicate power, and it is not wrong for a slave to reject that power relationship.

When people use it for hardware and software, there's a tacit understanding that the power relationship must be preserved.

These two are diametrically opposed.

> In slavery, the only correct thing for a master to do is to abdicate power, and it is not wrong for a slave to reject that power relationship.

Once that happens, they are no longer master and slave, of course.

In my field (FPGAs) master/slave is used correctly fairly often - having one module command another module is really common in HDL design. Although I'm happy to see "master/slave" gone, I haven't yet seen a good widely-adopted alternative.

For example, "primary/secondary" is slowly being adopted, but I find it very confusing because it can also mean two modules that perform the same function in parallel, but have a different priority for receiving input.

leader/follower, or commander/soldier could work. Maybe king/pawn or queen/pawn if we really want to uphold the same kind of relationship as slave/master.

Yup, after some thinking I've come to quite like commander/worker.

Some quick research seems to indicate that IDE "master" (Drive 0) and "slave" (Drive 1) operate in a true master/slave fashion.

There are times when the slave drive communicates with the master (rather than the host controller), and the master controls things.

I'm the opposite of an expert, but here's one example according to a document (https://www.ele.uri.edu/courses/ele408/s2001/projects/roland...) that I found. From section 6.3.4:

    During power on initialization or after RESET- is negated, DASP- shall be
    asserted by Drive 1 within 400 msec to indicate that Drive 1 is present.
    Drive 0 shall allow up to 450 msec for Drive 1 to assert DASP-. If Drive 1
    is not present, Drive 0 may assert DASP- to drive an activity LED.

There are a few other sections like 6.3.13 (power-on diagnostics) and 6.3.16 (optional spindle sync) where it's pretty clear that Drive 1 follows Drive 0's lead.

For cars with hydraulic brakes or a clutch, there is the master cylinder and the slave cylinder. the master cylinder is a pump that provides hydraulic pressure to the slave cylinder to do work.

I suggest the terms "boss" and "employee".

I know this is not HN custom, but this is gold. Bravo!

Hadoop?