[COFF] The MOO problem and set-uid

[COFF] The MOO problem and set-uid

[dave]

I recall reading in an old manpage that the (patented) set-uid bit was to 
solved the MOO problem.  I've searched around, but cannot find anything 
relevant.  Anyone know?

-- Dave

[COFF] The MOO problem and set-uid

[bakul]

The orig. ref. was in the Unix Time Sharing paper. See Section 3.5 near the end.
https://people.eecs.berkeley.edu/~brewer/cs262/unix.pdf <https://people.eecs.berkeley.edu/~brewer/cs262/unix.pdf>

The paper referenced above is behind a paywall (you can see the first page *free of charge*!)
https://onlinelibrary.wiley.com/doi/abs/10.1002/spe.4380010210 <https://onlinelibrary.wiley.com/doi/abs/10.1002/spe.4380010210>

One place the problem is described in some detail:
https://www.cs.unm.edu/~cris/481/gordon-moo.txt <https://www.cs.unm.edu/~cris/481/gordon-moo.txt>:


> On Jan 28, 2020, at 2:18 PM, Dave Horsfall <dave at horsfall.org> wrote:
> 
> I recall reading in an old manpage that the (patented) set-uid bit was to solved the MOO problem.  I've searched around, but cannot find anything relevant.  Anyone know?
> 
> -- Dave
> _______________________________________________
> COFF mailing list
> COFF at minnie.tuhs.org
> https://minnie.tuhs.org/cgi-bin/mailman/listinfo/coff

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/coff/attachments/20200128/08746f69/attachment.html>

[COFF] The MOO problem and set-uid

[dot]

Bakul Shah <bakul at bitblocks.com> wrote:
>
> The paper referenced above is behind a paywall (you can see the first
> page *free of charge*!)
> https://onlinelibrary.wiley.com/doi/abs/10.1002/spe.4380010210

That article mentions Frank King and Barry Landy who are still around in
Cambridge.

> One place the problem is described in some detail:
> https://www.cs.unm.edu/~cris/481/gordon-moo.txt

A correction: Titan was not a Unix system :-) It was a variant of the
Ferranti Atlas, with its own timesharing operating system. But it has some
second-hand Unix connections: it was the first system to have hashed
passwords, which Unix later adopted. Stephen Bourne used Titan when
developing Algol 68C.

Tony.
-- 
f.anthony.n.finch  <dot at dotat.at>  http://dotat.at/
Forties, Cromarty: West backing southwest later, 5 to 7. Moderate or rough.
Occasional rain. Good, occasionally moderate.

[COFF] The MOO problem and set-uid

[dave]

Thanks, all, for the responses; it was driving me nuts!  I know the game 
as "cows and bulls" (or perhaps that's because I'm of British origin).

I like Dennis' observation: "On the other hand, suid these days seems to 
be a fairly blunt and dangerous instrument".  I've often believed that if 
you thought you needed set-uid (esp. to root!) you can probably get away 
with set-gid instead.

-- Dave

[COFF] The MOO problem and set-uid

[bakul]

On Jan 30, 2020, at 12:25 PM, Dave Horsfall <dave at horsfall.org> wrote:
> 
> Thanks, all, for the responses; it was driving me nuts!  I know the game as "cows and bulls" (or perhaps that's because I'm of British origin).
> 
> I like Dennis' observation: "On the other hand, suid these days seems to be a fairly blunt and dangerous instrument".  I've often believed that if you thought you needed set-uid (esp. to root!) you can probably get away with set-gid instead.

Capabilities[1] (Dennis, Van Horn, 1966) would have solved the MOO problem.
And the CAP computer project had already started in Cambridge!

[1] 
https://www.princeton.edu/~rblee/ELE572Papers/Fall04Readings/ProgramSemantics_DennisvanHorn.pdf

Re-reading this papers decades later it is interesting to see that
the modern object caps are basically not very different from the
original concept! Also interesting to see fork() here.