quoted

The last time we discussed this (Andrew, Pavel and I), we've decided
that for ethernet PHY controlled LEDs we want the devicename part
should be something like
   phyN  or  ethphyN  or  ethernet-phyN
with N a number unique for every PHY (a simple atomically increased
integer for every ethernet PHY).

We might want to rethink this. PHYs typically have 2 or 3 LEDs. So we
want a way to indicate which LED of a PHY it is. So i suspect we will
want something like

ethphyN-led0, ethphyN-led1, ethphyN-led2.

I would also suggest N starts at 42, in order to make it clear it is a
made up arbitrary number, it has no meaning other than it is
unique. What we don't want is people thinking ethphy0-led0 has
anything to do with eth0.

Why do we have to distiguish between LEDs connected to the PHY and LEDs
connected to the MAC at all? Why not just name it ethN either if its behind
the PHY or the MAC? Does it really matter from the users POV?

quoted

I confess that I am growing a little frustrated here, because there
seems to be no optimal solution with given constraints and no official
consensus for a suboptimal yet acceptable solution.

I do think it is clear that the base name is mostly irrelevant and not
going to be used in any meaningful way. You are unlikely to access
these LEDs via /sys/class/leds. You are going to go into
/sys/class/net/<ifname> and then either follow the device symlink, or
the phydev symlink and look for LEDs there. And then only the -ledM
part of the name might be useful. Since the name is mostly
meaningless, we should just decide and move on.

Even more if it is not relevant ;)

-michael