| THINGS TO DO: |
| |
| + All device drivers must adhere to the DDI scheme of being able |
| to get configured dynamically (i.e. at run time). This requires |
| lots of work in the drivers, and in the way the "dev" structures |
| are handled. Before Linux 1.0 if possible. |
| |
| + Replace the ugly 'sk_buff' structure with the 'mbuf' structure, |
| and redo all code accordingly. |
| |
| + Move all device drivers to the 'drv' layer. |
| |
| + Impose a much stricter layering between the protocol modules. |
| |
| + Implement IP fragmentation. |
| |
| + Add support for the PPP and AX.25 protocols, possibly via the |
| external 'pkt' driver. This is NOT the current 'packet.c' module, |
| but a scheme in which we can implement external (user-space) drivers |
| for networking protocols, like SLIP, PPP and AX.25. |
| |
| + Add support for the Novell IPX/SPX protocols. |
| |
| REMAINING KNOWN BUGS AND PROBLEMS: |
| |
| + Local-port weirdness when using the 'r' utilities? |
| All connections seem to be using the same local (privileged) TCP |
| port (1023), and this cannot be the way God wanted it to be... |
| |
| + Sudden lockups when overloading a socket? |
| This seems to occur with X11 sessions (as per Linus Torvalds, 05/28/93) |
| and has its origin in the new timer.c code... |