RE: [MiNT] Security again

[Jo-Even.Skarstein@gjensidige.no]

> -----Original Message-----
> From: Joerg_Westheide@su.maus.de [mailto:Joerg_Westheide@su.maus.de]
> Sent: Wednesday, December 15, 1999 1:41 AM
> To: mint@fishpool.com
> Subject: [MiNT] Security again
> 
> JE>Well, is it any better to make it dependant on MiNT? 
> Atleast TraPatch
> JE>will work with TOS and MagiC as well.
> It won't be dependant on MiNT because the TSR will still be 
> available with a
> compatible (i.e. the new) interface

Then why not just use the TSR under MiNT as well??

> Yes, many people don't program for MiNT in first place.
> And I'm sure that most of them are not willing to create a 
> second version of
> their programs for MiNT. If it's necessary to bend a vector 
> under MagiC or
> TOS, they will do so under MiNT as well. And if you do not 
> allow vector
> bending under MiNT, you will just exclude this program from 
> running under
> MiNT.
> So the only thing you get is that you increase the splitting 
> of the Atari
> world.
> Is this what you want?

Who has said anything about denying vector-bending? The only way to do this
is to deny access to Supexec() or Super(), you can (AFAIK) alreay do this
today by selecting the correct SECURELEVEL-mode.

> JE>I'm not quite sure what you mean.
> He only suggested solutions for some existing TSRs.

(Just so there's no misunderstanging here - it was me who suggested these)

> His suggestion contained no information how programers could 
> avoid vector
> bending in their software

It certainly did, but there will always be cases where somebody feels that
vector-bending is the only solution - I'm 99% sure that somebody eventually
will find a need to circumvent TraPatch too.

> JE>Is there something that you can't implement as a XFS or XDD today?
> JE>Remember, you can bend vectors as much as you like inside 
> the kernel,
> JE>so basically anything can be implemented as a XDD.
> Where is the difference between a XDD/XFS and a TSR 
> concerning vector bending
> and stability?

The most serious problem with TSRs and stability (apart from bugs in the TSR
itself) is that the TSR can be damaged by other processes writing to it's
memory-space. XDDs are under complete control of the kernel. You are
allocated kernel-memory. The interface between the "TSR" and
user-application (if you have such a thing) goes through proper GEMDOS-calls
and not some cookie-based shared memory scheme.

Jo Even Skarstein