Correction to loopback-mode-type FACILITY and TERMINAL

[liulk]

https://openconfig.net/projects/models/schemadocs/yangdoc/openconfig-terminal-device.html

Currently:

• FACILITY: A port internal loopback at ASIC level.
• TERMINAL: A port external loopback at ASIC level.

Cisco reported that FACILITY and TERMINAL loopbacks should be reversed. @ahsaanyousaf confirmed it.

Proposed:

• FACILITY: A port external loopback at ASIC level.
• TERMINAL: A port internal loopback at ASIC level.

Background:

@aashaikh first added the definition here: https://github.com/openconfig/public/commit/4e0d236b04326c37275ea9d5c621ea51e7d8eeb8

@aashaikh first use of loopback-mode-type for terminal device loopback-mode leaf here: https://github.com/openconfig/public/commit/b05e0076a8716f4a21dd5029e22fc47d54b7d059

@sulrich changed the interface loopback-mode leaf from boolean to loopback-mode-type here: https://github.com/openconfig/public/pull/731

@sallylsy and @dplore added PHY and MAC specific enums: https://github.com/openconfig/public/pull/1034

[liulk]

cc: @aaronmillisor

[ahsaanyousaf]

cc: @ejbrever

[liulk]

Judging from the fact that https://github.com/openconfig/public/commit/4e0d236b04326c37275ea9d5c621ea51e7d8eeb8 and https://github.com/openconfig/public/commit/b05e0076a8716f4a21dd5029e22fc47d54b7d059 used terminology specific to optical transceivers, my guess is that it was originally only intended for PHY loopback, not MAC.

After #731 and #1034, it added cases where it is possible to mix the configuration of L1 loopback on L2 interfaces or L2 loopback on L1 channels.

• ASIC_PHY_LOCAL and ASIC_PHY_REMOTE (L1 loopback) on /interfaces/interface/config/loopback-mode which is an L2 interface.
• ASIC_MAC_LOCAL and ASIC_MAC_REMOTE (L2 loopback) on /terminal-device/logical-channels/channel/config/loopback-mode which is an L1 channel.

Do these actually make sense?

How do we feel about deprecating all enums of loopback-mode-type and just add LOCAL and REMOTE as the recommended values? The PHY / MAC will depend on which interface type the mode is configured on.

Updated proposal:

• Deprecate: FACILITY, TERMINAL, ASIC_PHY_LOCAL, ASIC_PHY_REMOTE, ASIC_MAC_LOCAL, ASIC_MAC_REMOTE.
• Add: LOCAL and REMOTE.

[LimeHat]

An ASIC can implement loopback functionality in the MAC layer or the (internal) PHY layer. There's also an external PHY layer which corresponds to a transceiver.

One of the points in #1034 was that in some use cases, the actual loopback functionality is not necessarily required and the FACILITY is used to emulate the oper-up state of an interface (when the operator doesn't care about the data plane behavior of the interface)

[ejbrever]

I agree the definitions of FACILITY and TERMINAL are reversed.

Maybe FACILITY|TERMINAL are not applicable to pluggable optics on a router the same way. In that case, perhaps dedicated enums for those would be better. But I would strongly suggest we do not deprecate FACILITY|TERMINAL as for optical networking they are industry standard names and we have a lot of OpenConfig implementations using those that we shouldn't break (at least ten across four vendors).