These are thing anyone can take w/o redoing things twice. Drop a line to
linux-streams@ordago.uc3m.es in case you pick one of them.

They are not in relevance order, just in casual order. Take a look to them all
and then decide.

You should also check ../NOTES, some things to be fixed are shown there and 
some things listed here in the past are shown there.

 ______________________________________________________________________

Check LiS code against SPEC, X/Open, POSIX, ... whatever
Check LiS code against Magic Garden, SVr4 man pages, Rago's book, ...

 ______________________________________________________________________

Implement file descriptor passing. Should be as simple as plug-in ioctl stuff
and little more. search lis_strioctl() for hooks to plug-in such code.

 ______________________________________________________________________

Write man pages for plain SVr4 STREAMS structs/syscalls/functions/ddi/dki/...

 ______________________________________________________________________

Complete streams tty stuff. Almost everything is there but you must check the
magic garden/whatever and ensure LiS head is doing the STREAMS tty stuff.

 ______________________________________________________________________

Revise priority band handling, again, the code is almost there but I'm sure
It's broken and there're some places where extra code should be added. Anyway
this should be just a minor change as the existing code is aware of priority
band stuff (u don't have to redesign anything at all).

 ______________________________________________________________________

Check I_ATMARK handling.

 _____________________________________________________________________

Answer the following questions ;-) They're mainly philosophical and perfomance
issues, but they could make a difference. BTW There're also ignorance caused
ones :-)

*) Where's best to put lis_runqueues() It's supposed to be called when any of
   (1) a streams system call may have caused a to-be-scheduled proc and (2) a
   ctx switch occur (just befor returning to user mode. But probably this is
   not the best solution. It'd slow down a little bit the whole system even
   when no streams processing takes place.

*) Where's best to insert a queue into scanq list? [Currently we assume a
plain FIFO is a good thing. but I guess this is not the best thing to do.

*) We use q_link for both scan & sched svc procedures, won't cause this
problems?

*) Is QSCAN flag needed (refer to queue.h)?

*) should we just putnext the held msg on lis_scantmout_handler or it's
actually better to call runqueues()

