[Info-vax] Future comparison of optimized VSI x86 compilers vs Linux compilers

[Arne Vajhøj]

On 8/3/2020 1:33 AM, Simon Clubley wrote:
> On 2020-07-31, Camiel Vanderhoeven <iamcamiel at gmail.com> wrote:
>> Op zaterdag 1 augustus 2020 00:05:56 UTC+2 schreef Stephen Hoffman:
>>> On 2020-07-31 20:53:43 +0000, Camiel Vanderhoeven said:
>>>> Op vrijdag 31 juli 2020 19:37:18 UTC+2 schreef Simon Clubley:
>>>>>  From a security point of view, RMS code is effectively kernel mode code
>>>>> anyway at the moment as you can get from executive mode to kernel mode
>>>>> without any additional privileges required.
>>>>
>>>> I'd wish you'd stop spreading this nonsense. Pardon my French. the
>>>> Exec/Kernel mode line was never intended as a security/privilege
>>>> related line, its a mechanism that helps stability by protecting kernel
>>>> mode data/code from bugs in exec mode, it's not meant to protect
>>>> against malicous code executing in exec mode.
>>>
>>> Nonsense?  The statement from Simon is correct, and then you (Camiel)
>>> then directly agreed with Simon's statement.
>>
>> I believe that an argument can be accurate, but that making it can
>> still be nonsense. Proper electrical grounding can help prevent
>> fires, yet I see very few people constantly harping on the fact
>> that it doesn't prevent arsonists from torching the place. Perhaps
>> nonsense is the wrong word here, I'll happily accept a better
>> alternative. >
> How about "inconvenient truth" instead ?
> 
>>> As you state, the rings are to isolate accidental corruptions and not
>>> to rebuff malicious activity.
>>
>> Yes, if only people could just accept that and move on.
> 
> It's one hell of a lot of additional complexity for relatively little gain.
> 
> Those additional modes could have been used to do _real_ isolation.

As stated the 4 modes was invented as a protection against coding bugs
not as a protection against hackers.

Does the 4 modes provide value? I think so.

With VMS design of expensive process creation and image activation
not starting a new process, then S mode is convenient to organize P1.

Does it make sense to have RMS data structures protected
against U mode write? Yes. Does it make sense to not have
RMS including index-sequential files which is really a NoSQL database
no running in K mode? Yes. So E mode does also make sense.

Would it have been better security if privs was necessary to
go S-E-K? Absolutely.

But that is what we can see in 2020. The world was different in 1977.

Remember 1977 was the year where Richard Stallman recommended
everyone at MIT to remove passwords to ensure anonymous access to
systems.

Hindsight is 20/20.

Arne