USB charging, part 1: requirements [LWN.net]

USB charging, part 1: requirements

Posted Jul 9, 2016 18:55 UTC (Sat) by marcH (subscriber, #57642)
Parent article: USB charging, part 1: requirements

> The other peer is the slave, or responder, that can only send or receive data when the host tells it to. This is known as a USB "device" though, since "device" is an overly generic term, it is sometimes referred to as a "gadget".

What was perceived wrong with USB "slave" that motivated the invention of "gadget"?

USB charging, part 1: requirements

Posted Jul 10, 2016 22:56 UTC (Sun) by anselm (subscriber, #2796) [Link] (9 responses)

There seems to be a general trend away from loaded terms like “master” and “slave” and towards more neutral terms. For example, what used to be called a “slave DNS server” is now often referred to as a “secondary DNS server”.

USB gadget vs slave

Posted Jul 11, 2016 0:10 UTC (Mon) by marcH (subscriber, #57642) [Link] (7 responses)

This "trend" crossed my mind only after I posted the question...

While the terms master and slave don't bring any good memories, they happen to have perfect technical accuracy. I wouldn't mind trading them against other terms, however it's going to be difficult to find alternatives which are both accurate and PC at the same time since the relationship between two pieces of hardware tends to be more brutal than touchy-feely. "Gadget" falls quite far off the mark.

From: flussence
> ... a pocket supercomputer into a brainless charging adaptor ...

Brainless power: see how the master/slave terminology is even more accurate in this particular case?

Anyway: while charging is an important and growing topic (thanks for the article), it's still not the bulk of the USB spec/complexity/implementation/etc.

USB gadget vs slave

Posted Jul 11, 2016 8:16 UTC (Mon) by farnz (subscriber, #17727) [Link] (2 responses)

Except that, in a deep and technical sense, the brainless power adapter is the USB "master" or "host", while the supercomputer attached to it is the "slave", "device", or "gadget".

And OTG makes it even more complex, as the "host" role can swap between devices.

USB gadget vs slave

Posted Jul 11, 2016 11:05 UTC (Mon) by tao (subscriber, #17563) [Link] (1 responses)

I think that was the whole point? That the slave owners are the brainless ones...

USB gadget vs slave

Posted Jul 11, 2016 13:13 UTC (Mon) by farnz (subscriber, #17727) [Link]

Trouble with that is that the terminology swaps around - the pocket computer is the device, whether it's attached to a desktop supercomputer host that outpaces it, or to a dumb charger host that couldn't outpace a 1970s home computer :)

USB gadget vs slave

Posted Jul 11, 2016 17:00 UTC (Mon) by mbunkus (subscriber, #87248) [Link] (3 responses)

> While the terms master and slave don't bring any good memories, they happen to have perfect technical accuracy.

I respectfully disagree with the term being technically accurate. For me the relationship between master & slave has two connotations: one is a question of rights (as in the slave doesn't have any and the master can order it/him/her around to do whatever) and one is who's doing the work (as in the slave does all the work and the master only assigns, supervises, punishes).

However, there are many situations where us admins speak about master/slave in technical environments and where those connotations don't really fit:

1. Databases. Here the master is often simply the single point where data can be changed. However, it's not like the master is not doing any work — often enough the work is divided somewhat equally between all nodes in the cluster.

2. High-availability clusters. Here the master is the one actually doing all the work, and it isn't even telling the slave what to do all the time. Usually the slave machine verifies that the master is still alive and takes over if that check fails — but that's usually not initiated by the master but by the slave's own will — and the nodes suddenly switch their roles, too. They're more often called active/passive clusters, and that's a much more fitting term to me; it's much more of a relationship between equal partners than any kind of hierarchy.

3. DNS servers. Again the masters may or may not do a significant amount of work depending on whether or not it's listed as a NS for the domain in question. Additionally it doesn't instruct the slave to update its files, it only notifies the slave. The slave decides based on its configuration if and when to update its zones. On top of that master and slave servers are often run by different entities (companies).

We use master/slave for so many different cases in tech that I'd prefer not to use them even without considering the historical context of those terms.

USB gadget vs slave

Posted Jul 11, 2016 17:13 UTC (Mon) by pizza (subscriber, #46) [Link] (1 responses)

> I respectfully disagree with the term being technically accurate. For me the relationship between master & slave has two connotations: one is a question of rights (as in the slave doesn't have any and the master can order it/him/her around to do whatever) and one is who's doing the work (as in the slave does all the work and the master only assigns, supervises, punishes).

There's a third connotation -- the question of authority. Ie the master is authoritative, the slave is not.

USB gadget vs slave

Posted Jul 11, 2016 21:37 UTC (Mon) by marcH (subscriber, #57642) [Link]

> There's a third connotation -- the question of authority. Ie the master is authoritative, the slave is not.

Yes - this is how this terminology has been successfully been used for decades.

Which side has authority and decides matters to correctness as in "Does it work?". Black/white questions[*]. Who does the work is optimization territory = quantitative stuff. Glad the terminology relates to the former.

[*] Sorry; really couldn't resist this one :-)

> > I respectfully disagree with the term being technically accurate.

"Send patches"

USB gadget vs slave

Posted Jul 14, 2016 1:24 UTC (Thu) by kevinm (guest, #69913) [Link]

These are some great points, and I would suggest that for all of these examples a pair of terms borrowed from partner dance would be appropriate: Leader / Follower. There's no connotation there of one doing more work than the other; there's the implication of "call and response" (another term from partner dance); and the possibility of switching roles as in the HA cluster case is apropos.

USB charging, part 1: requirements

Posted Jul 14, 2016 2:15 UTC (Thu) by gdt (subscriber, #6284) [Link]

The sentiment is correct but the DNS example is incorrect.

Mockapetris in RFC882 used "primary/secondary" to refer to the role of servers and "master/slave" to refer to the origin of DNS records. Terminology to distinguish the servers and the records they serve is perhaps too fine to be useful once the DNS system existed, whatever clarity the differing terminology bought during the design of the DNS system.

The use of primary/secondary in BIND4 configurations confused many. Master/slave was felt to be clearer and appeared in BIND8 configuration.

"Original" and "copy" — or many other suggestions — may have been better and perhaps clearer terms, but that would have required BIND to break away entirely from the terminology in the DNS RFCs.

Then there's the question of if inanimate objects such as servers should be anthropomorphised to the extent that a "slave server" is not a "neutral term"; a topic I am unqualified to determine.