Proposals like Lerner's "Segwit2MB" would activate the 2 MB hard fork after the activation of the Segwit soft fork. Maybe Lerner's 6 months (if I remember right) as "grace period" is too short because maybe until then most people would still use legacy transactions, but 1 year would be surely enough. We would then activate Segwit in mid-to-late 2017 and 2MB in mid-to-late 2018. Jihan as far as I remember mentioned the Hong Kong Agreement several times, so he should be happy with this solution.
segwit as a hardfork requires everyone to upgrade node.
but here is the thing.. native key users dont have to move funds to segwit keys.
what happens in reality is that instead of 2 merkle (block inside a block)..
its just 1 base block where the witness and txdata sit together.
and simple ALL share the same room.
its not
base 1mb || witness <3mb ||
[IN] [SIG] [OUT] || || - native
[IN] [OUT] ||[SIG] || - segwit
it is
base 2mb
[IN] [SIG] [OUT] || - native
[IN] [OUT][SIG] || - segwit
thus both tx types benefit and coexist in the same area with no stripping (filtering/bridging) data.. blocks are the same for all nodes of the network because they all upgraded to just the 2mb hardfork