OpenWrt 21.02.0 Fourth release candidate - DSA docs

[Paul D]

On 2021-09-17 13:27, Perry wrote:
> Hi all,
> 
> On 9/17/21 1:30 PM, Rich Brown wrote:
>> Hi Arınç
>>
>>> On Sep 17, 2021, at 3:17 AM, Arınç ÜNAL <arinc.unal at arinc9.com> wrote:
>>>
>>> The current naming used on LuCI/UCI is inaccurate and confusing. The “interfaces” under Network → Interfaces actually represent networks. The actual interfaces are called “device”.
>>
>> I agree that the terminology is confusing. I really struggled with the names when I added them into the preface to the DSA Mini-tutorial (https://openwrt.org/docs/guide-user/network/dsa/dsa-mini-tutorial). I did some research looking at the original DSA documentation: it didn't offer much in the way of definitions. So I followed my usual practice of documenting the lingo of whatever application I'm using.
>>
>> After looking hard at how LuCI seemed to work, I wrote:
>>
>> --------
>> 	• Devices are physical connections that convey bits/frames to other computers. They operate at layer 2 in the protocol stack, have a MAC address along with several other configurable parameters...
>> 		
>> 	• Interfaces route IP packets and operate at layer 3 in the protocol stack. An interface is associated with a single device that sends/receives its packets. Interfaces get their IP address parameters by the choice of protocol...
>> ---------
>>
>> I haven't heard any corrections from others about these assertions, so I am hopeful that I got those definitions right.
>>
>> When you say that "interfaces... actually represent networks" I think you mean that they're "subnets" (and have a subnet address range, IP address, and other characteristics). Is that what you mean? Although I'm neither a Linux OS or network expert, I can see an explanation for using the terms "devices" and "interfaces" as defined above.
> 
> This is not always the case.  For example, it is possible to have a tun
> or tap interface which does not have a corresponding ip address.  This
> is more than just a device, because layer 3 packets can arrive on such
> an interface.
> 
> Another example, from Freifunk, are mesh (either Ad-Hoc or 802.11s)
> interfaces.  These are interfaces which have a static IP address, but
> the netmask is 255.255.255.255.  This is not a network in the sense most
> people are used to using, but still a completely valid configuration.
> 
> I think staying with the terminology "device" and "interface" is the
> right way to go.
> 
> Greets,
> Perry
> 

Largely agree -  device --> hardware device --> hwdev

interface ....? GUI is an interface. NIC is an interface. This is 
abstract. Trying to make it concrete is just confusing. Specify:

L3interface
L4interface
MLinterface (multi layer)


This terminology desperately needs disambiguating (just an observation, 
not a criticism), at least away from single word terms.

But currently it seems to be:
-device is the hardware
-interface is hardware configured to operate at some network layer


What do Linux (kernel) devs use?