Another letter to TI

[Mychaela Falconia]

Hello FreeCalypso community,

I have just sent another outreach letter to TI:

https://www.freecalypso.org/members/falcon/TI-outreach/letter-20191011.pdf

I sent this letter to TI's Dallas headquarters, addressed to R. Gregory
Delagi from the leadership team:

http://www.ti.com/corp/docs/investor_relations/leadership.html

To be clear, my primary motivation for making these repeated outreach
attempts to TI is not because of the licensing nonsense, but because
of the missing bits: while we have recovered most of the source and
documentation materials related to the Calypso chipset, there are still
a few bits missing, and because of extreme obscurity, it is possible
that these last missing bits might not exist anywhere except TI's deep
archives, stashed behind a tall and solid wall of indifference.
Breaking through that wall of indifference at TI will probably require
a lot more than just a letter, what we really need is to have a crowd
of a hundred people camping out on a sidewalk in front of TI's HQ day
and night for a few months, demanding the release of those materials -
but we have to start somewhere, and the letter I just sent them is a
start.

The 10 source files (os_???.c and osx.c) which I asked for in the
current letter are not the only bits which we don't have, which may
exist in TI's deep archives and which would be highly valuable to the
mission of FreeCalypso.  There is the source code for the Calypso DSP
ROM, sources for Windows tools ccdgen.exe and str2ind.exe which we
currently limp without, schematics or other hw documentation for the
historical D-Sample board would also be very valuable (much richer
developer-oriented functionality than on our FCDEV3B), a copy of the
PADS PCB layout file for their quadband Leonardo board would be a very
useful additional reference - the list goes on, and this is just for
Calypso/2G.  In addition to their post-Calypso 2G chips (LoCosto and
E-Costo) which aren't too interesting to us, TI also had a working
solution for UMTS/3G - I personally knew a former TI engineer who
worked on it at their San Diego office, so I know their 3G solution
was working for real - but they threw it in the trash along with their
2G stuff when they exited that business because they felt they couldn't
compete against Qualcomm and MTK.

And no, I am not holding my breath for those os_???.c and osx.c files
I asked for in this letter - I will continue with my hardware plans
regardless of whether we ever recover those files or not - but I do
need to clarify to our FC community that I will not be able to work on
migrating from TI's proprietary compiler to gcc unless we obtain those
10 missing files.  We do have FC Selenite firmware which can be built
with gcc, using a not-too-trustworthy source reconstruction of OSL and
OSX from 2014 - but this fw is experimental, not production quality,
and it will remain so until and unless we recover those missing files.
It would be conceptually possible to do the compiler migration without
those original source files, working with disassembly-based
reconstruction only, but it would be an extremely unpleasant job, and
I am not currently motivated enough to take on a project direction of
such unpleasantness.  Thus as things stand, I will only be able to
produce gcc-built firmware that fully supplants the current TMS470-
built Magnetite version *if* we can recover those 10 missing files.

My FreeCalypso hardware work will continue no matter what, regardless
of whether we ever succeed in getting anything out of TI or not - as I
see it, I will spend the rest of my natural biological lifespan
(probably got some 30-40 y left) working two directions in parallel,
advancing FC hardware on one front (there is plenty more that can be
done on the hw front using the materials and know-how we already have)
and keeping up the social pressure on TI for the missing bits on the
other front.

However, there is also a place where the wider community can come in
very usefully, and this call to the community is the primary purpose
of the present announcement.  Breaking through the indifference at TI
and convincing them to dig up and release the ancient archival
materials we are after (as well as the whole license silliness) will
probably require a lot more than just my letters, we would also need
larger community pressure.  Specifically, if the letter I just sent to
TI fails to elicit any response just like my two previous letters sent
to their San Diego office produced no response, the proper next step
would be to amplify the demand with more people - it is easy to brush
off one person sending letters, but if we had a hundred people sending
them letters, that pressure would be more difficult to ignore.  The
present announcement is a heads-up to what is coming; if a couple of
months go by with no response from TI, then I will post again with
more ideas as to exactly how other community members can get involved
with further letter writing or other public pressure.

Meanwhile I keep working along my FC hardware roadmap.  On the hw
front my next coming-up project is MMTB1, a modem module test board
that will work with the already existing GTM900-B (and maybe GTM900-P)
modules, but is also expected to work with my proposed FCM40 module.
I already got the MMTB1 design done at the schem+bom level and got all
of the component footprints captured; the next step is the actual PCB
layout, which still remains to be done.  I will probably have MMTB1
done before my big surgery, but the more exciting FCM40 (the actual
ambition for which MMTB1 is a preparatory step) will probably have to
wait until after.

As for FCM40, my motivation there is two-fold: on the one hand it will
be our first hw product that is officially intended to serve as a
component to be integrated into larger systems (FCDEV3B is a development
board for lab bench use only, not for incorporation into larger
systems), and on the other hand it would be a perfect opportunity for
us to escape from the proprietary clutches of Altium.  The PCB data
model and ASCII file format of PADS are well-understood (at least by
me), and as soon as we have a FOSS PCB tool to migrate to (would need
to be a custom fork of either geda-pcb or pcb-rnd with some data model
changes), we have a ready migration path to move our OM-based modem
core PCB layout from PADS to FOSS.  But with Altium the situation is
much worse, i.e., the data model and file format of Altium are much
more hostile than PADS.

Because of the way things worked out back in 2015-2016, the PCB layout
work for FCDEV3B was done for us by certain people on a barter basis,
as opposed to a contractor hired with money.  Those people had moved
the design from PADS to Altium against my protests, and we've been
stuck with Altium for FCDEV3B ever since, held hostage to it.  Because
I don't have much hope for a migration path from Altium to FOSS, a
move to FOSS would require giving up all of the new layout work done
for FCDEV3B and going back to our OM GTA02 starting point, which is in
PADS format.  Throwing away our FCDEV3B PCB layout is not an option,
hence FCDEV3B remains stuck in Altium - but hopefully it won't need
any more changes beyond our current V2.

Now enter my FCM40 idea: that modem module won't need any of the
peripherals featured on the FCDEV3B, instead it will be just the modem
core (straight from GTA02) with just one FPC interface connector added.
That is a perfect opportunity to go back to GTA02 and to do the new
project right, do it without Altium, using only PADS and FOSS instead!
It is something I plan on working on in 2020-2021.

Finally what matters at the end of the day is this: even if we never
succeed in getting anything more out of TI than what we already have,
even if we never get a blob-free gcc-built fw version of production
quality, even if our current 98% source, 1.8% blob FC Magnetite fw
built with TI's ancient proprietary compiler is the best we ever have,
it is a still a quantum leap ahead of the competition!  Why?  Because
look at how unfree our competition is: our competitors (GSM/2G chipset
vendors) are MTK, Spreadtrum and RDA, and they are just as closed as
Qualcomm.  Just look at any of those MTK "source" leaks: *everything*
from the DSP-interfacing bottom of L1 to what they call L4 (seems to
be an approx equivalent of our ACI) is all binary objects, not one bit
of actual source for any of the interesting parts.  And it isn't just
the firmware, the available leaks in terms of chip register
documentation are just as bleak - nothing in comparison to the docs we
have for Calypso.

Therefore, unless MTK or Spreadtrum or RDA suddenly change their
stance and freely publish their complete chip documentation and
reference firmware source code (not to be expected obviously), we
still have a legitimate case for needing to produce and market our
FreeCalypso modems: despite the remaining few blob bits and despite
the lack of a FOSS license, our *published source* FreeCalypso
solution is still a monumental quantum improvement over the mainstream
MTK/Spreadtrum/RDA/Qualcomm status quo.  But if we can convince TI to
release those last missing bits which I am after, then our FreeCalypso
solution can be made even better and even more free, which is why we
need to continue a sustained campaign of public pressure on TI for as
long as it takes, however many years.

Hasta la Victoria, Siempre,
Mychaela aka The Mother