> Operating systems shouldn't be able to run unsigned software - ever.
We already have signed malware. This will only prevent you from installing your own self-compiled software or software from a trustworthy source. Few trustworthy sources will pay the money for a signature.
> ... proper, trusted certificate chain - no self-signed rubbish
Do you know how much a proper certificate costs? Also today we know that at least one attacker can control the root, and we have seen several CAs being taken over by non governments, as well as customers of CAs being issued certificates far wider than what they should have gotten. The CA world is a terrible mess.
> Certificate revocation lists should be enforced as strictly as is practicable.
Even Google now knows that revocation lists are bogus and possibly even harmful:
https://www.imperialviolet.org/2014/04/19/revchecking.html
> Sandboxing should be made to work properly, stricly enforced, ... as long as they've been given explicit permission by the end users
Look at the mobile world. People will enable _every_ permission they are presented with. As long as you cannot patch out the features in the source code.
It seems like you've never seen the discussions in the late 1990s where "Trusted Computing" came along which tried to do all of this.
The thing that actually did bring security since then was "Free and Open Source Software". FOSS scared Microsoft into (partially) cleaning up their mess they called Windows. Today when software crashes because of invalid input it's considered to be not just an unimportant bug, but a security problem which needs to be addressed immediately.