Idle: half-considering whether to using a UID/GID-style access-rights checking scheme (VUGID) in the BJX2 MMU (in addition to the usual User/Supervisor mechanism). Could add some additional memory-protection features without too much additional cost.
This isn't any more powerful than a normal MMU; it's just putting a fixed-permission-model hardware unit in place of software (kernel/supervisor) programming of PTEs/TLB to match whatever permission model they want.
-
-
It could allow permissions to be done on a per-thread basis though (by giving each its own "keyring"), and with less run-time cost than running all this logic through the PF or TLB miss handlers. Note that this would build on top of a traditional PTE/TLB system, not replace it.
-
Having multiple privilege domains sharing the whole virtual address space (as threads do), not just some mappings, is a huge security problem and inadvisable to try. Normal thread programming models don't do this; you use separate address spaces, not threads, to implement this.
-
This is also not intended to replace the use of separate address spaces for processes, but more for use within a process. Probably still better than a large application not having any protection internally (ex: the multi-decade issue of hostile plugins & ActiveX controls, ...).
-
You can't fix utterly wrong Windows architecture with a fancy MMU....
-
Variations seem to crop up every few years (often in different forms each time). This feature could better allow keeping plugins from sidestepping their API or 'sandbox'. Granted it is rather unlikely it would be adopted in a mainstream ISA. so mostly just an example here...
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.