Why would you need to allow modifications the binary und er LGPL? You will need to ship the source of qt and all mods you made to it, but that is it. Or am I missing sth?
Yes, you're missing the ability for the user to install and use the modified code, which is one of the main things that LGPLv3 introduced over LGPLv2.1.
Unless the software is written to ROM or otherwise made not modifiable by anyone, the creators of the GPL intend that the end user gains the right to modify this software. If using a version before GPL/LGPL/AGPLv3, it is ok if the proprietary parts of the system stop running if the software is not an official build. With the v3 versions, even this is not ok.
Now, I believe Linus Torvalds for example interprets the license differently, and doesn't believe that a company using Linux on an embedded system is obligated to allow their customers to apply a patch to this Linux system (and he certainly doesn't want the manufacturer to be forced to make their proprietary software run on top of the modified Linux, which is why he never wanted the Linux sources to be updated to GPLv3, even if it were feasible to do so).
With the GPLv3/AGPLv3 versions, it is still OK to break the proprietary parts. For the LGPL in general, you have to be able to re-link, so that isn't OK I think.
This is still a matter of interpretation. Here is an exchange that this same author has had with Stallman about Stallman's opinion on the GPLv3 (and his intention) [0]:
> Specifically, I asked him on 2012-05-05:
> > [so], these words in GPLv3: “The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made.” mean that the proprietary software that is not a combined work with the GPLv3'd work must also function?
> Stallman replied on 2012-05-06 with:
> > Absolutely. And I wrote it specifically to do that!
I don't think this has ever been litigated, so we don't know what a judge and jury would find, but I suspect that if they found Bradley Kuhn is right, there may be a GPLv4 that is more explicit about requiring the device to keep functioning as it was, not just as a dev board as he calls it in these slides. And that GPLv4 would apply automatically to lots of GPL software going forward (because of the "or later" clause that many projects are using).
I also think he is explicitly contradicting his earlier article to some extent. He also mentions in that that the whole original purpose of the GPL, as also told by Stallman, was to be able to fox your own printer by patching its code when you find a bug in it. It's clear that "turn the printer into a dev board by blowing a fuse" doesn't achieve this original goal at least - whether it achieves the license terms as written is more debatable.
OK, understood. One remaining question: is it enough to allow the re-linking to run in developer or emulated hw? Does LGPL e.g. forbid code signing of firmware?
No, according to the FSF and SFC at least. The goal is to allow end users to patch the GPL parts of their physical device. Whether the device has to keep functioning as before is more debatable, but being able to run the code in a separate, emulated, device is not good enough.
