Design of Design¶
Rational Model¶
Typical Process
Goal
Desiderata
Utility Function
Constraints
Resource Allocations, Budgets, and Crucial Budgets
Design Trees (with attribute and alternative branches)
The hardest part of design is deciding what to design.
A chief service of a designer is helping clients discover what they want designed.
We usually don’t know the design tree, we discover it as we go.
At each node of a design tree, one faces not a simple alternative choice among options for one design decision, but an alternative choice among multiple tentative complete designs.
The goodness function cannot be evaluated incrementally due to combinatorial explosion.
One must trim the design tree as one goes down via experience and estimators.
The desiderata and their weightings keep changing.
The constraints keep changing, which makes the Waterfall Model wrong and harmful.
Fighting requirements bloat and creep.
Appoint early-on strong, seasoned, domain-knowledgeable managers who can stay with the program through initial systems delivery.
Empower them to tailor standardized processes and procedures as they feel is necessary.
The use of a Requirements Traceability Matrix ensures that each requirement detailed, defined, and laid on is indeed derived from one or more of the initial overall requirements.
The definition of clear key performance parameters by Milestone A and clear requirements by Milestone B that can remain stable through Initial Operational Capacity can be essential to an efficient development phase.
The necessity for contracts best explains the persistence of the Waterfall Model for designing and building complex systems.
The Co-Evolution Model¶
Creative design is not a matter of first fixing the problem and then searching for a satisfactory solution concept.
Creative design is a matter of developing and refining together both the formulation of a problem and ideas for a solution, with constant iteration of analysis, synthesis, and evaluation processes between the problem space and the solution space.
Although this is a better model than the Rational Model, Fred Brooks does not think it is sufficient.
Hickling’s Whirligig Model is too complex with its cycles and epicycles.
Raymond’s Bazaar Model¶
Raymond argues that the system products produced by the bazaar process are in general technically superior to those produced by the cathedral process.
The market mechanism selects, in evolutionary fashion, the best-designed modules.
Subjecting a new module simultaneously to hundreds of testers smokes out the bugs sooner, yielding a more robust product.
Bugs are fixed better because of market selection among fixes.
When the bazaar will work (according to Brooks).
The results are not always quite polished or quite debugged; they only have to be good enough for the purpose of the builder.
An overall system design already exists.
A key to all design processes is the discovery of the users’ needs, wants, and criteria.
Boehm’s Spiral Model¶
Brooks suggest punctuating the spiral with explicit contracting points, augmented with clear specification of what can be contracted, with what certainties, and with what explicit distribution of risk.
Brooks advocate frequent but not continuous interaction with representative users, with successive prototypes as the vehicles.
The Rational Model has persisted in practice despite its inadequacies and plenty of cogent critiques because of its seductive logical simplicity, and that builders and clients need contracts.
Costs of Collaboration¶
Partitioning Cost
Learning/Teaching Cost
Communication Cost during Design
A mechanism for change control must be put into place so that each designer makes only those changes that affect only his part, or have been negotiated with the designers of the affected parts.
The Challenge is Conceptual Integrity¶
Properties of conceptual integrity: Coherence, Consistency, Uniformity of Style, Orthogonality, Propriety, and Generality.
How to achieve conceptual integrity with team design?
Negotiation among peers is the classic recipe for bloated products.
The result is design by committee where none dare say “No” to another’s suggestion.
Empower a single system architect.
The user interface must be tightly controlled by one mind.
Name the scarce resource explicitly, track it publicly, and control it firmly.
When Collaboration Helps¶
Determining needs and desiderata from stackholders.
Brainstorming
Competition as an alternative to collaboration, which could lead to unplanned design competitions resulting in product fights.
Multidisciplinary design review
When Collaboration Doesn’t Work for Design Itself¶
Each part of a design has at any time one owner.
Some resolution procedure must be in place for settling conflicts.
Some version control must be established so that each designs against a single time-stamped version of all the earlier design work.
If collaboration tools are designed so they encourage committee design, they will do more harm than good.
Even in the conceptual design stage, when conceptual integrity is most imperiled, pairs of designers acting uno amino can be more fruitful than solo designers.
Making Telecollaboration Work¶
Face-to-face time is crucial.
Clean interfaces.
Technologies for Telecollaboration: Document, Telephone, Telephone + Shared Document, and Videoconferencing.
Design Perspectives¶
Rationalism vs Empiricism
User Models
Better wrong than vague.
Write down exactly is known about the user, the user’s purposes of use, and the modes of use.
Write down what is assumed (guessed) about the users and the uses.
Constraints that shrink the designer’s search space.
Real constraints
Obsolete once-real constraints
Constraints misperceived as real
Intentional artificial constraints
Esthetics and Style in Technical Design¶
Structure, Metaphor, Consistency.
Orthogonality: Do not link what is independent.
Propriety: Do not introduce what is immaterial.
Generality: Do not restrict what is inherent.
Exemplars in Design¶
Exemplars provide safe models for new designs, implicit checklists of design tasks, warnings of potential mistakes, and launching pads for radical new designs.
Not laziness, but a high level enthusiam and diligence is required for the mastery of the corpus of exemplars available in any design domain.
He who seeks to make designs that really work is most apt to come up with new designs of enduring value, almost as a by-product.
The desire to be original has degraded many a work.
Study failure examples even more carefully than you study successes.
Watch yourself after success. Success stimulates confidence in the design technique, in the design itself, and in oneself. All may lead to overconfidence.
Think at the top level about the object you are designing and its assumptions about the environment in which it will be used. Is a paradigm shift under way? Will your assumptions still be valid a decade later? Are you designing the right thing?
Remedies for the Divorce of Design from Use and from Implementation¶
Use-scenario experience
Close interaction with users via incremental development and iterative delivery.
Designers need to dig more energetically and personally into the actual experiences and processes of implementation.
Design curricula simply must include both techniques for and practice at understanding users’ needs and desires.
Linearizing the Web of Knowledge¶
Cut edges in the graph until it is a tree to impose a hierarchical order where there was none before, whether that order is wanted or not.
Map the tree onto a line in any of the several well-known ways (e.g. depth-first).
Insights into the Design Process¶
Design isn’t just to satisfy requirements, but also to uncover requirements.
Design isn’t simply selecting from alternatives, but also realizing their existence.
The tree changes as the design changes, how to represent that evolution?
A decision tree with \(n\) independent binary design questions yields a tree of designs with \(2^n\) nodes.
Modular versus Tightly Integrated Designs
One may need to trade off quality of the design for speed and ease of the design process itself.
Modular designs are more readily represented as decision trees.
One sketches in the big units and then proceeds by progressive refinement, adjusting dimensions and adding more details.
Progressive truthfulness start with a model that is fully detailed but only resembles what is wanted. Then one adjusts one attribute after another brining the result ever closer to the mental vision of the new creation, or to the real properties of a real-world object.
The dream system starts with a rich library of fine specimens of fully detailed designs.
Starting with exemplars that themselves have consistency of style ensures that such consistency is the designer’s to lose.
The true hazards of progresive truthfulness lurk in the library.
Shortcomings (e.g. bad models, too few models, too narrow a variety of models) will limit the emerging designs.
Visual Displays - Multiple Concurrent Windows¶
The historical limitation of one time-shared window enforced the following unnatural and wasteful behavior:
Study a big spatial chunk for context.
Zoom in
Create or manipulate some local portion.
Zoom out
Repeat
Have two big screens showing a context view with a detailed view all the time.
The Workbook View¶
Yet another window concurrently displays the designer’s workbook.
Designers come to design stations with
In-progress designs.
Action plans they take away.
Updated in-progress designs and future action plans.
Logs capturing all actions enable automatic backtracking.
Notes, ideally dictated so as to leave the hands free to design, documenting what was tried and reasons for rejection, acceptance, etc…
Great Designs come from Great Designers¶
Product process is conservative, aiming to bring similar but somewhat different things within one orderly framework.
Aims at predictability.
Encouraging tactics that have worked in the past and discouraging those that have failed.
Veto-oriented aimed at blocking bad ideas and catching oversights.
Aims to inhibit a confused product line, where one’s own products are one’s most dealy competitors, and customers don’t know what to buy.
Demand consensus by many people and the past precedence.
Designed for follow-on products.
Capability Maturity Model¶
We’re at Level I [on the Capability Maturity Model] and will always be.
—Steve Jobs
How to make a process that encourages great designs?¶
The product procedure must explicitly identify the matters of fundamental importance and constrain those and only those.
All rules can be broken.
Go for conceptual integrity: entrust your design to a chief designer.
References
- BJ10
Frederick P Brooks Jr. The design of design: Essays from a computer scientist. Pearson Education, 2010.