The problem space defines dimensions of design improvement. These dimensions are:

Description: Durability measures the persistence in time of the ability to achieve the intended function. By this definition, durability covers such diverse criteria as reliability (MTBF) and the production of extremely short-duration events such as femtosecond laser pulses.

- friction (causes wear and heat)
- corrosion
- wear (of materials)
- cavitation pitting
- charge leakage (e.g. DRAM)
- photographic impermanence (cause? fading?)
- instability (of particles and configurations)
- incoherence (of quantum states)
- obsolescence (of devices)
- consumption (of resources)
- burn-out
- exceeding operating characteristics (duty cycle, etc)
- exceeding environmental characteristics (ambient temperature, etc)

Description: Efficiency measures the amount of resources consumed to achieve desired output Efficiency is usually measured as a percentage of some theoretical optimum. Perfect efficiency is ruled out by the second law of thermodynamics.

Efficiency is (at least) a second order problem in that it describes the ratio of desired output to resource consumption.

We divide resources into consumable and fixed (not consumed) resources to get a more complete picture. So a rocket consumes fuel but also has mass (weight). The efficiency of a rocket isn't simply how completely it turns fuel into kinetic energy. It must also take into account the mass of the rocket.

If we take into account the *cost* of the structure and fuel, we get a measure of the cost per pound of payload, which is a more practical measure of efficiency. Next, we account for the cost of labor, maintenance, etc over the entire life-cycle of the project. So thermodynamic efficiency is only a piece of the picture.

Technologies don't become practical until they reach some threshold of efficiency. The possibilities of a technology aren't even explored until that threshold is reached. E.g., electric power generation didn't take off until the Gramme Dynamo.

Corollary: Incremental improvements in efficiency almost always win out over radical redesigns for larger gains. (cf., silicon semiconductors vs. III-IV semiconductors, gasoline automobiles)

Conjecture: Conservation of efficiency - integration and automation often increase operating efficiency, but reduce maintainability and therefore ultimate durability. Total efficiency does not increase linearly with operating efficiency.

- unamortizable overhead
- modularity (opposite of function sharing)
- indirection (in representation, measurement & control)
- physical limits
- speed of light & other propagation speed limits
- thermodynamic limits
- I
^{2}R losses - drag increases as the cube of velocity.
- eddy currents
- friction
- thermal losses
- optical losses (e.g., photovoltaics, fiber optics)
- signal degradation
- incompressibility of liquids, etc
- cognitive limits
- perceptual limits
- attention
- psychomotor limits

- mathematical limits
- np-completeness
- combinatorics
- stacking & packing limits
- topological, geometrical, algebraic, arithmetic constraints e.g. non-planarity, network flow, etc.

Description: Flexibility measures the ability to adapt a design to achieve a different function. E.g., a truck is more flexible than a train.

*Why design for flexibility?*

- rapid adaptation to changing environment,
- reuse of design effort for new problems.

- function sharing (opposite of modularity)
- limited information
- narrow restrictions on environmental conditions
- autocatalysis or lock-in (e.g. software that is version-specific)
- hard-coding constants (software)
- non-standard parts (cf., Eli Whitney & replaceable rifle parts)
- functional fixity (c.f. Dunker) - cognitive blinders

Description: Generality measures the narrowness or specificity of the function to be achieved. (applicability) E.g., a programmable computer is more general than a fixed function calculator. A material is more general than a device.

*Why design for generality?*

- bigger markets
- more utility

- functional fixedness
- limited information processing capacity
- limitations on speed of adaptation
- tradeoffs with efficiency (e.g. Swiss army knife - bulky & inconvenient)
- cost of reconfiguring
- cost of replication

Description: Power measures the adequacy with which the design achieves the function. It characterizes the scaling of the output in terms of work per unit time, peak speed, accuracy, etc.

- difficulty of containment (at high temperatures/field strengths)
- strength of materials
- inefficiencies at high power (e.g., rockets must lift themselves and fuel)
- inefficient use of resources

- cost of fabrication (e.g., cost of semiconductor fab houses)
- imperfections and inaccuracies in materials

Description: One of the great success stories of 20 century technology was the tremendous increase in computing power for a given cost. This has been due largely to miniaturization, the ability to implement increasingly more circuitry in ever smaller packages. Scaling down a technology often makes it economical. On the other hand, economies of scale often arise from scaling up, in order to amortize the fixed costs of manufacture and generation. In so far as natural laws and mathematical properties don't all scale uniformly, barriers to scalability form one of the main obstacles to design and technology improvement.

The other scaling truism is in science, the smaller the object you want to observe, the more
energy (and larger machine) you need to look at it. Particle accelerators are *big*.

- non-linearities in physical laws i2r losses drag increases as the cube of velocity Stefan-Boltzman law : total radiation from a black body is proportional to the 4thpower of its absolute temperature.
- non-linearities in geometry
- radius/surface/volume
- surface-area effects:
- heat transfer
- diffusion
- evaporation

- different physical laws applicable at different scales (e.g. quantum effects)
- scarcity / limited resources
- inefficient or impure building blocks
- bottlenecks and non-parallelizable processes