Can we state a MT switch running swos is a L2 switch while the same switch running ros is a L3 one instead ??
No, simply running ROS on device doesn’t make it L3 switch. It’s the configuration that makes the difference - IP routing in particular (L3 is IP). SwOS can not do any of L3 functions, ROS can (if configured so). But it is quite easy to configure ROS to only provide L2 functionality.
Keep in mind that the OSI 7-layer model doesn’t apply precisely to TCP/IP, intended as it was to describe a competing network model that never really took off. Your average web browser covers layers 4 thru 7, for instance, whereas in the old OSI model, you’d need four different program layers to do it “right.”
This mismatch gives marketing people freedom to play games in this area. For instance, multicast is up at the protocol layer, so there are “smart” switches with barely enough brains to do IGMP snooping that get called “L3 switches.”
As mkx indicates, the difference comes when routing, which is why MikroTik can justify the moniker “Cloud Router Switch.” Yes, it’s a router, and it’s a switch. This line of products are better at switching than routing, but the fact that your CRS series switch can also route packets between subnets suffices to make it a true L3 switch. Note that the SwOS-only devices are confined to the “CSS” scheme; they don’t route packets at all.
When you restart one of the dual-boot CRS devices into SwOS, you do lose nearly all the features of what we’d call an L3 switch.
I’d say that for a switch to be called “a true L3 switch” it has to be able to do L3 (routing) at wire speed.
Today there are some MikroTik switches that can do that using HW-accelerated routing, although the software is in beta phase. Many older MikroTik CRS devices were not able to do it because they could only do routing via the CPU.
However, other “true L3 switches” on the market have been able to do this for many years, and their firmware is perfectly stable.
Thanks for clarifying.