Lists Home |
Date Index |
I accept it. Not amazingly, the success factors look a lot
like the best practices for web services. Boltzman's
Theorem still is the best predictor. By the way, the enemy
of the lessons learned and presented in this article is the
dictum, "if it can't be measured, it can't be managed". It
makes the Seibel-style vendor very nervous when you explain
to them that a software project can't be managed by "a price
and a part number".
"Low-precision architecture diagrams are easier to remember, low precision
requirements tables are easier to prioritize and evaluate early in a project,
low-precision design documentation is better at giving the reader "the idea"
of the design - and then letting them look around."
Why a basic schema is often better than a large complex set of state diagrams. Only a computer really
likes extremely detailed instructions. People find them tedious, insulting, and expensive. The
problem we discovered with CASE systems for enterprise modeling is that we confused
modeling with simulation.
"Some people like to make lists, some don't. Some work best nights, some work
best in the morning. Some like deadlines, some don't. Groups vary similarly. Some
cultures prize public self-scrutiny, others shelter people from embarrassment, and so on."
You have to model a culture if you want to model human communications.
Simple process/control models a la IDEF0 really do work but don't go
"Physical proximity. I am at a loss to explain why, but being physically
close to the other person affects the communication. Whether it is
three-dimensionality, timing, smell, or small visual cues, physical proximity matters."
That he is at a loss to explain it tells me he was a hacker
and probably had a lot of trouble getting a date in high school.
"Multiple modalities. People communicate through gestures as well as words, often making a
point by gesturing, raising an eyebrow or pointing while speaking."
God gave you at least five senses. Learning how to turn them off is the first
skill hackers learn so they can ignore their mother screaming at them from
the kitchen to turn off the computer and come to dinner.
"Vocal inflection and timing. By speeding up, slowing down, pausing, or changing tones,
the speaker emphasizes the importance of a sentence, or perhaps its surprise value. "
Unfortunately, through years of practice, screaming becomes the only vocal the
modern programmer hears through the headphones and the grungy rock.
"Real-time question-and-answer. Questions reveal the ambiguity in the speaker's
explanation, or the way in which the explanation misses the listener's background.
The timing of the questions sets up a pattern of communication between the parties. "
"Yeah." "Right." "Uh huh." "Didja?" "No, YAGNI!" and so it goes.
Proxemics, haptics, kinesics, culture, gestures, sensory modalities and timing.
Score for the HumanML guys. It seems semiotics
and hermaneutics still have the most accurate models of human communications
and computer scientists have the worst.
I want to publicly thank Ed Dumbhill for the HyTime award for HumanML. It honors
us, reminds us to keep the hype minimal and because it keeps people away from the project,
just might enable it to succeed. ;-)
From: Marcus Carr [mailto:firstname.lastname@example.org]
This is getting even further off-topic, but Rick Jelliffe sent me this pointer
the other day. It reflects the way that he manages the developers of Topologi
products and if you accept the basic premise, it casts quite a different light
on the use and importance of methodologies. Even if you don't accept it, at the
very least it illustrates the difficulty and imprecision of picking a
development strategy for a project. See