Proposal for the WLAN mesh routing

[Ivan Krstić]

Andi wrote:
> I have not found a "white paper" about the Marvell mesh protocol and its
> in depth function principles.

The mesh protocol is not Marvell's; it's an implementation of the
802.11s draft.

> With b.a.t.m.a.n, you could have a working routing daemon *right* *now*.
> And chances are good that b.a.t.m.a.n could be implemented in the
> Marvell driver in the near future.

The Marvell blob is getting replaced by a fully free firmware currently
in development. As with the Marvell FW, however, routing has to happen
on-chip in order for us to maintain the ability to serve as a mesh
router even when the computer is suspended. Therefore, anything that
doesn't fit the *tiny* ROM+RAM available on the mesh chip simply doesn't
work. You're not talking about userspace or kernel execution here;
you're talking about executing directly within a RTOS on a little ARM
processor. Needless to say, existing code is very difficult to port to
this environment.

-- 
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | GPG: 0x147C722D