Re: [stack] Concatenative Hardware
"Christopher Diggins" <[email protected]>
| Newsgroups | gmane.comp.lang.concatenative |
|---|---|
| Message-ID | <[email protected]> |
On 11/2/07, William Tanksley, Jr <[email protected]> wrote: > Christopher Diggins <[email protected]> wrote: > >>for some time: Could you build a machine with > >>capability-based addressing? That is, instead > >>of building an MMU (not that you said you were > >>going to) could you protect a process' memory > >>from other processes with capabilities? > > >Yes I could do that, but I am currently studying > >minimalist system architectures. > > Then you'll certainly want to study Capability based security. A > capability is a pointer without which it is impossible to access a > given object. For non-mission critical embedded systems, is this really a huge gain? I can imagine though it being useful for embedded devices on cars and rockets, and other such things. How should an architecture react in the case of catastrophe (i.e. capability violation)? Some kind of interrupt mechanism I imagine? > Very powerful, very simple. Shapiro (KeyKOS, EROS, Coyotos) has > written a lot about it online. The paper > http://citeseer.ist.psu.edu/689826.html was useful to me. I'll look at it soon. > The Baker model I pointed to would make Capabilities very simple to > build; it would be entirely impossible (not merely improbable) to > access a data structure which isn't linked to from something already > given to you. The kind of architecture I am considering at this time is so tiny that the overheard of capability pointers would be too big. However, as soon as I move into larger architectures I can see advantages. > > Would there be commercial interest in this line > > of exploration? Right > > about now some kind of financial sponsorship > > would be very motivating. ;-) > > Grin. > > I don't know, but I doubt it. > > Unless maybe you can convince Chuck Moore that your work would go > perfectly on his SeaForth24 chip. He's got plenty of money now that > Intel and AMD are paying him royalties on his patents. You know, that isn't a half-bad idea! It would be interesting to implement a Cat machine using one of those SeaForth 24 chips. - Christopher