I don't think "primary" CPU is a good name. We should call it application processor like is common in the smartphone world, because that is what it is. The other processors should not be called "secondary" either, simply because they are secondary in performance. What matters is which part controls which other part.
To your position, I think there are some caveats to note: One is when the non-free firmware does not run on the "primary" (application) processor, but runs concurrently on another processor and can subvert what is running on it (such as DMA capabilities of Intel ME, or baseband firmware in some smartphones).
Also if you offload non-trivial computation to e.g. your GPU then its firmware becomes relevant too.
Basically I found the positions of all sides involved hypocritical, since the original discussion around Atheros Madwifi HAL - which is basically just what other Wifi products ran in firmware. How are you any more or less free depending on whether the HAL is executed on your "primary" CPU, or inside some ARM processor in your Wifi card?
> If non-free firmware is a threat to user freedom then allowing it to exist in ROM doesn't do anything to solve that problem.
I get the argument for that somewhat. If even the vendor cannot update, then that mitigates user subjugation to some extent, because you can at least not force new versions on users. And FSF says this also applies to firmware that is in principle updateable but in practice this largely doesn't happen (the FPGA programming in NVIDIA G-Sync monitors comes to mind).
Some caveats
To your position, I think there are some caveats to note: One is when the non-free firmware does not run on the "primary" (application) processor, but runs concurrently on another processor and can subvert what is running on it (such as DMA capabilities of Intel ME, or baseband firmware in some smartphones).
Also if you offload non-trivial computation to e.g. your GPU then its firmware becomes relevant too.
Basically I found the positions of all sides involved hypocritical, since the original discussion around Atheros Madwifi HAL - which is basically just what other Wifi products ran in firmware. How are you any more or less free depending on whether the HAL is executed on your "primary" CPU, or inside some ARM processor in your Wifi card?
> If non-free firmware is a threat to user freedom then allowing it to exist in ROM doesn't do anything to solve that problem.
I get the argument for that somewhat. If even the vendor cannot update, then that mitigates user subjugation to some extent, because you can at least not force new versions on users. And FSF says this also applies to firmware that is in principle updateable but in practice this largely doesn't happen (the FPGA programming in NVIDIA G-Sync monitors comes to mind).