Re: Crypto Bounties: Another Thought that crossed my mind.

snow@smoke.suba.com said:
>       Well, I was thinking, what if a "Crypto Software Bounty Server"
>  were set up, so that someone could propose a tool that they would like
>  to see, along with an initial bounty. Others could contribute toward that
>  bounty (anonymously if they wish) until either the tool was delivered.
> 
>       The original issuer sets standards for the software (i.e. "easy to
>  use interface to mixmaster remailers for Macintosh", then must define
>  easy to use; Software considered delivered when in [alpha beta late-beta
>  &etc.]). The first to present software meeting these qualifications gets
>  the bounty, with the caviate that the software must be either gnu-copylefted,
>  or some similar "free use" copyright, after all, "The Net" paid for it...

Hmmm.  This is a one shot game (is that the term?) whereas software
generally has implications that escape a single sale scenario.  For
example, the more difficult the software, the more risk there is that
someone else will beat you, thus lowering the real risk-adjusted payoff
dramatically.   For this reason, more complex stuff would need some sort
of contract+reputation scenario that allows a repeating game to work.

A contractual alternative could work like this.  I (the initial desirer)
write a contract specifying my requirements.  I publish this as a market
tender, where other desirers can contribute funds, and this becomes a
cumulative sum that grows.  Programmers can offer to supply by naming a
price.  The market clears when the pot of desire equals or exceeds the
lowest offer to supply.

In a simplistic scenario, the contract is now sealed and the work is
done and paid for.

In order to overcome project failure, I could write my contract as a
multi-supplier seed project (often done by governments).  That is, the
pot gets shared around, say, the three best alternatives.  Once
supplied, all are free to pick from the alternatives.

In order to overcome the low silly bid, somehow reputation would have to
be built into the market.  That is, your efforts in the past as
programmer will cause your solution to be better valued than mine.

So perhaps we should turn this around.  Programmers would write
contracts to supply the given requirements at clearance+T, also
specifying the clearance price.  Your price is 2000 for the widget, mine
is 1000.  (That makes three contracts, a widget spec, my work plan and
your work plan.)  The market (the desirers) would then push funds back
and forth between the two until it became clear that one or the other
cleared.  As more pressure increases, more funds pour in.

Then, deliverables could be phased, with monies similarly phased, so
that the market has a chance to monitor.  Now, if it is not delivered,
reputation suffers, and you have to lower your price for the next job
(or hide in shame).

There's a lot of aspects of newbies and switching funds that I havn't
really thought through here.  However, I like this viewpoint because it
eliminates the need for judges.  History shows that a good market
microstructure will beat an authority approach in the long run.

Also note that if you drop the free software assumption, and make it,
say, moneyware, then the market becomes much more workable - the asset
being traded is a share of future revenues.  This has more ramifications
than might be obvious:  Propose a market to write a GAK killer for
e$10,000.  If it clears and is built, is the Dept. of Justice forced to
buy the rights out?

> Has anything like this been proposed before?

I don't know if this has been proposed before, but it is a logical
conclusion of the CP world.  If the Internet transaction costs make
single-person economic entities the most efficient unit in the machine,
and e$ in all its forms provides the oil, then we need some way to
connect up the parts for grand projects where there are hard cash
incentives.  There has to be some mechanism to distribute that cash in
the most efficient manner to produce the result.

BTW, this is a here and now issue, not a hypothetical future.  We have
already found that there is too much work on our plate, and limited
efficiency in farming it out.  Others may have found the same.

-- 
iang
iang@systemics.com

> snow@smoke.suba.com said:
> >       Well, I was thinking, what if a "Crypto Software Bounty Server"
> >  were set up, so that someone could propose a tool that they would like
> >  to see, along with an initial bounty. Others could contribute toward that
> >  bounty (anonymously if they wish) until either the tool was delivered.
> >       The original issuer sets standards for the software (i.e. "easy to
> >  use interface to mixmaster remailers for Macintosh", then must define
> >  easy to use; Software considered delivered when in [alpha beta late-beta
> >  &etc.]). The first to present software meeting these qualifications gets
> >  the bounty, with the caviate that the software must be either gnu-copylefted,
> >  or some similar "free use" copyright, after all, "The Net" paid for it...
> Hmmm.  This is a one shot game (is that the term?) whereas software

     Correct term, but not necessarily accurate. 

> generally has implications that escape a single sale scenario.  For
> example, the more difficult the software, the more risk there is that
> someone else will beat you, thus lowering the real risk-adjusted payoff
> dramatically.   For this reason, more complex stuff would need some sort
> of contract+reputation scenario that allows a repeating game to work.

     Not necessarily. It could be set up to keep the Bounty open, to provide
a revenue stream for upgrades, or second, third & etc. bounties for upgrades
bug fixes and new features. 

> A contractual alternative could work like this.  I (the initial desirer)
> write a contract specifying my requirements.  I publish this as a market
> tender, where other desirers can contribute funds, and this becomes a

     <snip>

> eliminates the need for judges.  History shows that a good market
> microstructure will beat an authority approach in the long run.

     Reputation could also eleminate the need for judges. If Matt Blaze, 
or Randall S. were to try to claim a specific bounty, people would be 
more likely to accept their claim than if I were to do so. 

> Also note that if you drop the free software assumption, and make it,
> say, moneyware, then the market becomes much more workable - the asset
> being traded is a share of future revenues.  This has more ramifications
> than might be obvious:  Propose a market to write a GAK killer for
> e$10,000.  If it clears and is built, is the Dept. of Justice forced to
> buy the rights out?

    If it is GNU'd, then there is no one to buy out. I like the idea of 
"free" software. As it stands now, there is a lot of very high quality 
software free software out there (hell, all of the software I use _daily_
is "free" software), and if no one owns it, the government can't use 
copyright laws to restrict it. The feds _can't_ buy it out for a million 
dollars. 

     The other thing about the "bounty server" is that it is simpler to
set up (at least as far as I can see) than your contract model. 

     All you would need is to settle up the details, set up a web server
with the ability to handle ecash, and some sort of accounting and you are
off. Maybe, just to prove you are legit, a performance bond. 


Petro, Christopher C.
petro@suba.com <prefered for any non-list stuff>
snow@smoke.suba.com

At 6:05 am -0500 11/19/96, snow wrote:
>     Reputation could also eleminate the need for judges. If Matt Blaze,
>or Randall S. were to try to claim a specific bounty, people would be
>more likely to accept their claim than if I were to do so.

But, what about reputation distortions?

There's the famous story about a guy presenting his discovery of the normal
distribution to a scientific society. Gauss was in the audience. He stood
up and said something like, "Oh. I figured that out years ago."

We now call it a "Gaussian" distribution, among other things.

If I remember the story correctly, Gauss never subsequently proved that he
discovered the normal distribution first, and he certainly never published
it at the time he said he discovered it.

Nobody remembers the name of the guy who was presenting the paper, though.
At least, I can't now. :-).

I'm not saying that Gauss *didn't* discover the normal distribution. I'm
saying that he didn't have to *prove* he did. Of course not. He was the
greatest mathematician of his time, and probably since.

I'd call the event a reputation distortion.

Cheers,
Bob Hettinga

-----------------
Robert Hettinga (rah@shipwright.com)
e$, 44 Farquhar Street, Boston, MA 02131 USA
"The cost of anything is the foregone alternative" -- Walter Johnson
The e$ Home Page: http://www.vmeng.com/rah/

> At 6:05 am -0500 11/19/96, snow wrote:
> >     Reputation could also eleminate the need for judges. If Matt Blaze,
> >or Randall S. were to try to claim a specific bounty, people would be
> >more likely to accept their claim than if I were to do so.
> I'm not saying that Gauss *didn't* discover the normal distribution. I'm
> saying that he didn't have to *prove* he did. Of course not. He was the
> greatest mathematician of his time, and probably since.
> I'd call the event a reputation distortion.

      With either system proposed (market based contracts v.s. bounty) 
you are paid for code. Let's face it, with reputation capital, losses (i.e.
bad moves/actions/whatever) are far more costly than good moves pay. 

     To use a sports analogy: If you fumble the ball, and allow the runner
to score, more people are going to remember it than if you make a couple of
baskets.

     What was the thing that killed Bushes chances of re-election against 
a rather weak canidate? One lie "No new taxes".

     If Gauss had been called on it, what would have happened? If the caller
could _prove_ he was lying, what then? He still would have been the
greatist mathmatician of the time, but he would have been seen as a liar
and a crackpot. We know how that works don't we.  


Petro, Christopher C.
petro@suba.com <prefered for any non-list stuff>
snow@smoke.suba.com

snow wrote:
> 
> > I'm not saying that Gauss *didn't* discover the normal distribution. 
> > I'm saying that he didn't have to *prove* he did. Of course not. He 
> > was the greatest mathematician of his time, and probably since.
> > I'd call the event a reputation distortion.[...]
>      If Gauss had been called on it, what would have happened? If the 
> caller could _prove_ he was lying, what then? He still would have been 
> the greatist mathmatician of the time, but he would have been seen as 
> a liar and a crackpot. We know how that works don't we.

No.

I think the more common case is "the rich get richer, the poor get 
poorer." The truth is insufficent when honesty puts you at a 
disadvantage.

-rich