Python obfuscation

Alex Martelli aleax at mail.comcast.net
Sun Nov 20 03:26:11 CET 2005


Anton Vredegoor <anton.vredegoor at gmail.com> wrote:
   ...
> Suppose I grant all your theories about optimal marketing strategies.
> This still doesn't account for the way the market is behaving *now*. It
> isn't in any way logical or optimal. For example in Holland (where I
> live) complete governmental departments are dedicated to make life

What makes you think that governmental departments are part of the
*market*?!  Government behavior is controlled by laws that are vastly
different from those controlling market behavior; if you're interested,
you could study the "theory of public choice".

Studying "perfect" markets (which can be mathematically proven to be
optimal in some senses -- the Arrow-Debreu Model, for example) is
parallel to studying physical systems that do not have attrition or
other such complications -- it's mathematically sharp (not easy, but WAY
easier than taking account of all the complications of the real world),
intellectually fascinating, AND practically useful in many (not all)
cases, since many real systems can be usefully modeled as "perfect" ones
with "perturbations" (second-order effects) considered separately.

If Galileo had tried to START physics by studying real-world systems in
all of their complexity, we'd still be at square zero; fortunately, he
was able to identify systems "close enough to perfect" (e.g., heavy
weights faling to the ground, compact enough to ignore air resistance,
etc) to get the ball rolling.  Physics still faces a lot of challenges
after many centuries in areas where the "perturbations" are much
stronger than "second-order" -- I'm told our physical modeling of cloud
systems or such everyday phenomena as welding is still way from perfect,
for example (forget quantum and relativistic effects... I'm talking of
everyday observations!-), and of course so does the much younger
discipline of mathematical economics.  Nevertheless the study of the
"perturbations" is well under way, with Nobel memorial prizes having
already been awarded in such fields as "bounded rationality" and
asymmetric-information markets.

Just like a gunner in the mid-19th century had to know fundamental
physics pretty well, but also developed experience-based heuristics to
compensate for all of the "perturbations" in his system, so the
practicing economic actor today needs a mix of science and art (in the
original meaning of "art", of course).


> You seem to tackle the problem of python obfuscation by first proving
> that it isn't feasible and then giving some kind of solution that will
> work and give the desired result: webservices. However when I look at

That seems to me to be a practicable technical approach, yes.

> obfuscation techniques I see a desire to profit from giving some person
> the idea that he or she is superior to someone else because he has a
> better product. In order to avoid copying we now need obfuscation. The

You're discussing the *motivation* for obfuscating, while what I was
giving was a possible way of *implementing* similar effects.

> difficulty to copy the thing (whether it is a swiss watch, a sportscar,
> designer clothes, the latest computer game, an ipod, a computer
> program) is part of the advertising game and is the basis for
> associating it with a certain status. If you look for a few minutes at

Yes, this is close to the theory of luxury goods (which is WAY
underdeveloped, by the way: if you're a post-doctoral student in
economics and are wondering where best to direct your research in order
to stand a chance to gain a Nobel memorial prize some day, you could do
worse than turn your efforts to this field).  The maths are complicated,
in the theory of luxury goods, because utility is RELATIVE: buyer's
advantage cannot be computed, even theoretically, from knowing just the
buyer's utility curve and the amount of good supplied to that buyer,
because the curve depends on the *relative* amounts supplied to that
buyer versus other buyers.

Throw asymmetric information into the mix, and, besides a mathematically
unmanageable mess, you get the famous anomaly whereby an INCREASE i the
price may locally result in an INCREASE in demand (backwards-sloping
price-demand curve) -- each buyer, lacking information about other
buyers, infers it from price signals, and *assumes* (as is normally the
case) that higher price means fewer buyers; since fewer buyers means a
higher relative advantage, this assumption increases each buyer's
appetite and thus, in the aggregate, raises demand.

While this is fascinating as an open research topic, AND crucial to
economic survival for purveyors of luxury good, I dispute the wisdom of
trying to model MOST markets of interest as subject to the fascinating
complications of "luxury goods theory".

Take, again, the specific example of the sawmill with NC saws, doing
custom work by cutting customer-specified shapes out of standard planks
of wood.  If things can be optimized, so that six such shapes can be cut
out of each plant rather than the five which would result from a simple
layout of the shapes, the mill can meet an order for 3000 shapes by
consuming 500 planks of wood, rather than taking up 600 planks.  There
is no need at all for any of the complications of luxury-goods theory to
understand and model this: the developer of the superior heuristic has
created "objective" value, under the simple and natural assumptions that
wood costs money and wasting less wood for the same output is therefore
an indisputable savings.

NOW you may get into the issue of how that value is split between
supplier (of the heuristic) and buyer (the sawmill), under various
possible arrangements.  Market segmentation may well enter the picture
here, because there may be different orders of widely different sizes --
maybe 3000 shapes is a typical order, but there may be some for 300
shapes and some for as many as 30,000; if the supplier charges the same
price for usage of his heuristic for any size of order, either the
heuristic will not get used for the smaller orders (making the overall
"pie" of value to share smaller), or the buyer will capture close to all
the value for larger orders ("buyer's advantage" situation).  So, it is
definitely to the supplier's advantage, and it can be shown that
situation exists in which it's of MUTUAL advantage, if different prices
can be used for the same (good of) service when used in different
situations (size of orders), which is exactly what market segmentation
is all about.


> a TV screen and notice what the commercials are trying to tell you, you
> will see that it's almost always that you will be better, stronger,
> more popular or beautyfull etc. if only you use product X.

Whether such ads WORK, of course, is an entirely open question; targeted
ads, which get paid for only when they DO work, appear to be the
direction advertising is taking these days.

Anyway, ads are practically _irrelevant_ to all we were talking about so
far, except in as much as they may help in asymmetric information
markets, which is a separate (and infinitely fascinating) sector of
economic theory.  It seems to me that you're mixing in all sort of
somewhat anecdotal observations, without a sound basis in either
economical theory or deep practical experience, just as when you were
taking your opinions about government departments as somehow related to
the working of markets (?).

If Galilei had started dispersing his energies and attention by worrying
about the COLORS in which the balls he was dropping to the ground were
painted, rather than focusing on RELEVANT issues such as size and mass,
he'd hardly have made much progress on the issue, would he?-)


> You are perfectly right if you would say that it is an illogical
> strategy to make people feel better relative to other people in order
> to get them to do something you want. Commercial entities could in

I'm not saying that: if I had to sell luxury goods, I would definitely
pursue such a strategy.  However, there are plenty of goods and services
which do NOT particularly need the complications of luxury-goods theory.

> principle be free of such things but we live in a world that is
> dominated by this worldview and if one tries to sell something one has
> to take that into account.

If you're selling luxury goods, sure, you can't afford to ignore related
issues.  But for most goods and services, particularly in the "business
to business" sector, the approach (which DOES get tried, as you can see
by some ads in magazines such as the Economist, Business Week, Forbes,
and so on) is IMHO quite silly (and a good example of that 50% of money
spent on advertising that is proverbially wasted).


> So how to get the same kind of market segmentation (as you call it)
> when you deploy your program as a webservice and where essentially the
> cost for you (moving a few electrons to produce a solution to a
> problem) is exactly the same whether you give the user a good or a bad
> result. If you give optimal results to everyone, users will go to other
> sites just because these sites give them opportunity to feel better
> than other people, not because this is objectively better, but just
> because that is how they think the world "works".

I believe that you are quite wrong: if your results are in fact better
than other sites', users will come get your results.  Markets are
imperfect, rationality is bounded, etc, etc, but users in general are
not total morons, to deliberately go and get worse results "to feel
better than other people".  For example...:

I was reading recently that Google's market share of web searches has
grown from 47% to 57%, comparing September 2004 with Sep 2005, for
example.  If your theory had ANY validity whatsoever, this should be
impossible, since Google does give the best results it can to any comer;
therefore, it would follow from your theory, another site could steal
our traffic by serving artificially-degraded results to the non-paying
public, and true search results only to subscriber, or something.  The
world just does not work this way -- thanks be.

Similarly: until recently, Opera "degraded" the user experience of
non-fee-paying users (by devoting a portion of its window to showing
banner ads) and required user to pay a fee (so, according to your
theory, "felling better than other people") to get a pure, undegraded
browsing experience.  Firefox came along and "gave optimal results to
everyone" instead.  Result: Firefox ate Opera's lunch, forcing Opera to
change its business model drastically.  These events, once again, are
totally incompatible with the theory you advance in this paragraph.

I doubt you will be able to do a good job of comprehending, much less
_explaining_, market segmentation strategies and tactics, unless you
take the trouble to shed the ideological baggage, and stop trying to
force a vision of the huge system that's the complex of markets in this
world through the narrow slit of that small subset of those which are
"luxury-good markets".  I don't deny that luxury goods exist (that would
be just as silly as your attempt to claim that ALL goods and services
fall under the complexity of luxury-good market theory!): I specifically
claim that luxury-goods situations are a small subset of markets (they
may be very visible, and can command large profit margins to offset
their small volumes and large advertising costs, but they're still
NICHES compared to "where the action is", namely ALL OTHER markets put
together).


>  Alternatively, lets just forget about obfuscation and try to get
> people to freely share by promoting open source (and open webservices).

I tend to a more pragmatic stance, just like Eric Raymond: although open
source is going to prove preferable in _most_ cases, there _will_ be
other cases (like the NC saw example, where ESR himself was the
consultant who advised the inventors to keep their new heuristic a
secret) where the overall production of value in the world (quite apart
from who's going to capture what fraction of that value) will increase
if certain kinds of innovations can be exploited by their inventor.

Since redistribution of value, as long as a lot of value is created, can
be dealt with by other means, maximizing the creation of value tends to
be the goal I prefer -- a policy that quashes part or all of value
creation based on redistributive precepts is, by this very fact, going
to be something I look askance at (making the pie smaller to try to
ensure that what little is left gets sliced according to your political
preferences, rather than ensuring the pie is as big as possible as the
first order of business, and dealing with the slicing issues as
_secondary_ ones).


Alex



More information about the Python-list mailing list