Those who cannot remember the past are condemned to repeat it. –George Santayana The past is never dead. It’s not even past. –William Faulkner Most developers pursue the Latest and Greatest with intense fervor, yet the history of engineering, including software projects, contains rich lessons that we risk repeating ad nauseam. This session recounts a variety of stories of projects that failed…and why. Ranging from the Vasa in 1628 to Knight Capital in 2012, each story tells of a mistaken interpretation of some architectural fundamental principle and the consequences–some good, some less so. I also look at the common threads for these stories, which resonates with problems many companies have but don’t realize. Pets. Com, San Francisco project, The Vasa, F16, Sagrada Familia, Tacoma Narrows Bridge, null, Ada, Serialization, Knight Capital, Ariane 5, Webvan, Chandler project

1. Stories Every Developer
Should Know
@neal4d
nealford.com

3. Vasa

7. 72x24lb/11Kg

12. 2-
deck ships
cannon
ships

13. 2-
deck ships
cannon
ships
2-deck cannon ships

14. #scaling_architecture

15. some things in
architecture
don’t scale linearly

16. when restructuring architecture…
…reevaluate “—ilities”.

17. requirements

18. requirements + “—ilities”

19. defining architecture

25. no !

26. #tradeoffs

27. Tacoma Narrows
Bridge

30. By Tysto - Self-published work by Tysto, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=477955

33. #scaling_architecture

34. F16

35. 1974

37. requirements + “—ilities”
#feasibility

38. Why?

39. null

40. https://www.infoq.com/presentations/Null-References-The-Billion-Dollar-Mistake-Tony-Hoare

45. https://www.infoq.com/presentations/Null-References-The-Billion-Dollar-Mistake-Tony-Hoare

46. Person x = new Person();
Cat y = new Cat();
x = y; //not permitted

47. x = ?

48. x = ?

49. ?

50. Algol

51. Algol

54. why we can’t have
nice things
#legacy

55. #reuse

57. serializationserialization

58. #implications

59. Ada

61. Lady Ada Lovelace

62. Ada

63. Ada
HOLWG

64. Ada

65. Ada

66. Ada

67. > 450
1983
37
1996
DOD Languages Used

68. COTS

69. #ubuiltit
#uownit

70. #standardization

71. #overengineering
(#forced)

72. Ariane 5

75. 0110101101011010
0110101101011010011010110101101001101011010110100110101101011010
guidance system (16 bit):
flight data (64 bit):

76. 01101011010110100110101101011010011010110101101001101011010110100110101101011010
guidance system (16 bit):
flight data (64 bit):
Some of those bits might have been important!

77. guidance system (16 bit):
v5 faster than v4
#reuse

78. guidance system (16 bit):
only used on the ground…
#debuggingNproduction
…kept on for first 40s of flight…

79. #debuggingNproduction
#legacy
#abstractiondistraction

80. https://www.thoughtworks.com/insights/blog/implications-tech-stack-complexity-executives
increased tech
stack complexity
#abstractiondistraction

81. https://www.thoughtworks.com/insights/blog/implications-tech-stack-complexity-executives
increased tech
stack complexity
#abstractiondistraction

82. https://www.thoughtworks.com/insights/blog/implications-tech-stack-complexity-executives
#abstractiondistraction

83. #debuggingNproduction
#legacy
#abstractiondistraction
an int is an int…

84. pets.com

87. #toolittleinfrastructure

88. webvan

90. why webvan failed

91. why webvan failed

92. why webvan failed

93. #toomuchinfrastructure

94. #toomuchinfrastructure
Let’s build all the
frameworks we’ll need first!

95. #misunderstanding”—ilities"

96. Knight Capital

97. “bankrupt in 45 minutes”
dougseven.com/2014/04/17/knightmare-a-devops-cautionary-tale/

98. SMARS
PowerPeg

99. http://www.nanex.net/aqck2/3522.html

101. SMARS
PowerPeg

102. SMARS
PowerPeg

103. #doyouevenDevOps?

104. clean up technical debt
#reusegonebad

105. San Francisco
Project

106. Every business is pretty
much the same, isn’t it?
spoken by someone who has never
implemented a real-world business solution

110. auto and
homeowners
insurance
division
commercial
insurance
division
casualty
insurance
division
life
insurance
division
disability
insurance
division
travel
insurance
division
customer
customer
customer
customer
customer
customer
customer
service

111. #canonicality

113. bounded context ≠ entity

114. The more reusable something is,
the less usable it is.
— Evolutionary Architectures
http://evolutionaryarchitecture.com

115. Microservice is a label, not a
description.
— Martin Fowler
https://martinfowler.com/articles/microservices.html

116. San Francisco Foundation and Utilities
San Francisco Common Business Objects

117. San Francisco Foundation and Utilities
San Francisco Common Business Objects

119. #last10%rule

120. “Users always want 100% of what they
want (& are never satisfied with less).”
80%
what the user wants
10% 10%
#last10%rule

121. #last10%rule
#reuse

122. Chandler
Project

123. infinite time
infinite resources
no legacy

124. https://www.amazon.com/Dreaming-Code-Programmers-Transcendent-Software/dp/1400082471

126. By OSAF - http://en.wikipedia.org/wiki/Image:Chandler_calendar_screenshot.png, CC BY 3.0, https://commons.wikimedia.org/w/index.php?curid=3876130

127. We’ve consistently overinvested
in infrastructure and design, the
fruits of which won’t be realized
in the next development cycle or
even two—that is, not in the
next six or twelve months.

128. I’m more and more feeling like
the art here is to do agile
development without losing the
long-term vision—and, frankly, I
didn’t even define the problem
as that to start with.

129. #metawork
is more interesting than work.

131. #metawork
is more interesting than work.

132. Sagrada
Famila

142. #experimental
LMAX

143. Vasa
Tacoma Narrows
F 16
null
Ada
Ariane 5
pets.com
webvan
Knight Capital
SF Project
Chandler
Sagrada Familia

144. #tradeoffs
#scaling_architecture
#feasibility
#legacy
#standardization
#reuse
#toolittle
#toomuch
#doyouDevOps
#canonicality
#metawork
#experimental
Those who do not learn
history are doomed to repeat it.
—George Santayana

145. #tradeoffs
Vasa
F 16
Ada
SF Project
Chandler
null

146. #tradeoffs
#scaling_architecture
#feasibility
#legacy
#standardization
#reuse
#toolittle
#toomuch
#doyouDevOps
#canonicality
#metawork
#experimental
Those who do not learn
history are doomed to repeat it.
—George Santayana

147. #metawork
Envelopes
webvan
SF Project
Chandler
Ada

148. Vasa
Tacoma Narrows
F 16
null
Ada
Ariane 5
pets.com
webvan
Knight Capital
SF Project
Chandler
Sagrada Familia
#reuse

149. #reuse
design
Vasa

150. #reuse
design
Tacoma Narrows

151. #reuse
old code
null

152. #reuse
everything
Ada

153. #reuse
old code
Ariane 5

154. #reuse
not enough!
pets.com

155. #reuse
invest too early
webvan.com

156. #reuse
old toggles
Knight Capital

157. #reuse
∞San Francisco Project

158. #reuse
design
Chandler
code
frameworks
platforms

159. message bus
process choreographer
service orchestrator
business services BS BS BS BS BS BS
enterprise services ES ES ES ES ES ES
application services AS infrastructure services IS
ESB-SOA

160. Customer
Code Reuse (Over Time)
checkout shipping
ESB-SOA

161. checkout shippingmicroservices
Code Reuse (Over Time)

162. The more reusable code is,
the less usable it is.

163. checkout shippingmicroservices
Code Reuse (Over Time)
service
component
module
module

164. service
component
module
module
service
component
module
module

165. service templates
service
component
module
module
logging monitoring
A&A
http://www.dropwizard.io/https://projects.spring.io/spring-boot/

166. #reuse
carefully

167. #tradeoffs
#scaling_architecture
#reuse
#toolittle
#toomuch
#doyouDevOps
#canonicality
#metawork
#experimental
Those who do not learn
history are doomed to repeat it.
—George Santayana
The past is never dead.
It’s not even past.
—William Faulkner
#feasibility
#legacy
#standardization

168. nealford.com
@neal4d
nealford.com/books
nealford.com/videos
www.oreilly.com/software-architecture-video-training-series.html