> On June 3, 2015 at 9:33 PM John Morris <jmorris@???> wrote:
>
>
> On Thu, 2015-06-04 at 02:52 +0200, Adam Borowski wrote:
> > On Wed, Jun 03, 2015 at 06:18:37PM -0500, John Morris wrote:
> > > Non-free software: NO, Firmware: YES. So ixnay on things like the Nvidia
> > > drivers but yes on blobs. The reasoning on where to draw the line is
> > > pretty clear cut.
> >
> > How exactly firmware is not software? Both are strings of bits encoding
> > commands for a processor living in silicon you own.
>
> So if the manufacturer puts the same firmware in an eeprom it isn't a
> problem? Or the BIOS itself? Are you running a Free BIOS? Do YOU know
> what your ACPI BIOS is doing right now? How about the CPU, ...
...
> In a more perfect world I'd agree that all that stuff should be open
> too, but it ain't, it ain't going to be. RMS managed to find -one-
> oddball machine that meets his definition of Free, if the vendor of that
> machine tried to sell them on the open market outside China they would
> find few takers. Bunnie's Novena 'Open Laptop' has blobs and closed 3d
> video drivers as well. Good luck tilting at this windmill.
>
> Where we can and should draw the line is in the kernel's address space.
> Blobs loaded into the kernel make the entire system untrustworthy and
> unmaintainable in ways a firmware blob loaded at initialization into an
> entirely different microcontroller managing WiFi doesn't. Not to
> mention that for regulatory reasons most vendors just aren't going to
> discuss the point with us. The situation stinks but changing it is
> beyond our current capabilities.

For some years RMS used a Lemote Yeeloong notebook with a 10 inch screen (the
oddball machine you were referring to). Several years ago the continuing
availability of that machine became doubtful. Around the same time it was
discovered that a certain model of Thinkpad could be corebooted and had
acceptable freeness, so he switched to that.

Earlier this year another computer was brought on the market:
http://www.fsf.org/resources/hw/endorsement/gluglug
(I don't know if RMS is using this one or not). http://www.fsf.org/ryf and
https://www.h-node.org/ are useful resources for finding hardware supported by
free software distributions.

Last I heard, RMS applies the following criterion to firmware: if the user can't
change it, it doesn't have to be examined for freeness.

So, the power controller on the Yeeloong was exempt because it can't be changed.

But the Raspberry Pi relies on a video controller blob which is loaded at boot
time and won't function without it. So it is not free because the blob could in
principle be changed.

I don't know what he thinks about chips which can be programmed using JTAG but
which don't get firmware loaded at run time.

And as far as FPGAs in general are concerned I don't know of any which can be
programmed without the aid of proprietary software tools and secret data stream
formats. I would like to be proven wrong about this. Recently I heard about
Cubic ( http://cubicboard.org/ ) and OpenCores ( http://opencores.org/ ) has
been around for a while.

Peter Olson