> Does this need to be built into Mir directly? Could it be a helper library
> that is Mir data structure aware?
What advantage would there be to separating this into a new helper library? It shares code with other functions in libmirclient.
> If it serializes settings, it is logical to expect that not just the shell
> would want to edit those settings.
It is logical that serialization is available to the same code that can (already) edit display configurations through the client API.
> I'm also interested in the human readability of this blob. I'd rather not see
> us introducing yet another binary format, especially one based on an
> internally-used library. It could be JSON or XML or something simple.
I'd not realised human readability was a requirement, OTOH parsing JSON or XML would introduce a new dependency.
> Does this need to be built into Mir directly? Could it be a helper library
> that is Mir data structure aware?
What advantage would there be to separating this into a new helper library? It shares code with other functions in libmirclient.
> If it serializes settings, it is logical to expect that not just the shell
> would want to edit those settings.
It is logical that serialization is available to the same code that can (already) edit display configurations through the client API.
> I'm also interested in the human readability of this blob. I'd rather not see
> us introducing yet another binary format, especially one based on an
> internally-used library. It could be JSON or XML or something simple.
I'd not realised human readability was a requirement, OTOH parsing JSON or XML would introduce a new dependency.