[mpls-linux-general] Confusion on status

[Jonathan D. <im...@ya...>]

Hi,

I have to admit a degree of confusion regarding the current status of MPLS on Linux and the current direction. I'm hoping that someone can shed light on the following:

1) People appear to be running the existing code, that much is clear from postings to the list, but what exactly is the current status of the code? For example, what's the ratio of "old" code to code that's aware of kernel changes since the last major release?

2) There's lots of routers that would likely benefit from MPLS-awareness, but is the current thinking that this should go into each software router independently (eg: BIRD, Quagga, XORP, Click, etc) or to have it in the kernel with virtual circuits exposed as logical devices so that there's one implementation for all?

3) There are plenty of networking projects that involve MPLS in some way (Sourceforge alone has several) and others that are not really tied to MPLS but would likely be used by people using MPLS. If Linux-MPLS is below critical-mass, would it make sense to cobble together some sort of alliance between projects to get momentum and interest up a bit?

4) I saw comments along the lines of other OS kernels being more interesting at the moment, but by-and-large the differences between kernels are abstraction details. The protocol doesn't change, the logic of passing packets is independent of the exact struct used for a packet (or indeed an interface), so unless the code is going to be rewritten as an netfilter/nftables module, you've two entirely distinct problems - interfacing with the OS and doing something useful. (Which is why portable software routers are possible at all.) Does the change in OS direction indicate that the split between the problems needs revisiting?