S?bastien Bourdeauducq wrote:
> No no. We keep everything on the internal flash. I do not want
> involvements with low-level debugging of the memory card protocol
> idiosyncrasies (and there are many)

Hmm, what specifically are you concerned about ? If all you
want is read blocks from a memory card and don't care about
optimizing speed or power consumption, you should be able to
ignore pretty much all those weird descriptors and things.

I mean, you've already implemented memory card support. What's
missing for a simple boot loader ?

> coupled with users deleting the
> wrong files or formatting the card with unsupported filesystems.

Oh, you want filesystems :-) Well, FAT16 probably won't go away
too quickly. (And there's no need for patent-encumbered VFAT.)

Partition tables may face extincion sooner, i.e., when memory
cards approach 2 TB. Probably sooner, because magnetic disks
already pass this barrier.

If you're really worried about that, you could always fall back
to a sequential search, allowing people to write their boot
file with "dd" and whatever the legacy operating systems use.

> Also, there might be no memory card and only Ethernet in the end.
> That would even be preferable if possible without blowing up the
> cost.

Hmm, that doesn't sound very nice from an integration point of
view. The memory card makes the device self-contained. If it
depends on Ether, you need a rather complex life support system.

> I do not believe USB storage will ever happen.

Why not ? If you target this low-cost device specifically at
developers, it wouldn't be too unreasonable to expect that some
of them will actually contribute to the infrastructure.

Or did you mean "patch developers" ?

> I do not
> expect USB stack functionality to grow a lot because this work is
> painful and no one wants to do it.

I think it all depends on getting the critter to run Linux
properly and then turning that USB host into at least a
general Full-Speed host (i.e., without offloading some of the
USB stack into the Navre).

I agree that most people able to make significant contributions
to that side of the system would consider extending RTEMS' USB
stack a dead end.

> >It's not quite clear to me how you want to position the device.
> >If it's for M1 developers, they'd probably want more M1 features.
> >If it's intended as a cost-down replacement of M1, that may work.
> >If it's for FPGA enthusiasts, you can probably cut off some more
> >fat.
> 
> It is an entry-level M1, which is all of the above.

Heh, now you have conflicting feature sets :-)

Maybe we have to go back a few steps. What is it really that's
bothering you most about the current situation ?

- Werner
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode

Reply via email to