Hipster Tech, Cargo Culting, and you being of no use (to me)

I gave this talk at Railscamp 18 in Canberra, 11th–14th of December 2015.

Transcript

1. Hipster Tech, Cargo Culting, and you being of no use (to me).
2. Sebastian von Conrad @vonconrad
8. Cargo
9. Cult
10. cult |kʌlt| noun a religion or sect considered to be false, unorthodox, or extremist, with members often living outside of conventional society under the direction of a charismatic leader.
11. cult |kʌlt| noun a religion or sect considered to be false, unorthodox, or extremist, with members often living outside of conventional society under the direction of a charismatic leader.
12. cult |kʌlt| noun a religion or sect considered to be false, unorthodox, or extremist, with members often living outside of conventional society under the direction of a charismatic leader.
13. under the direction of a charismatic leader
15. Rails is the best and worst thing that happened to Ruby.
16. Let me explain...
17. Rails propelled the popularity of Ruby...
18. ...but Rails itself was too popular.
19. We all want to build web apps quickly.
20. And Rails seemed like a good way to build web apps quickly.
21. Build web apps quickly Cargo
22. Cult Rails
23. But while Rails allows rapid development, what about maintainability?
24. History has proven this to be (generally) false.
25. RubyConf AU 2015 (aka SOA Conf)
26. People who have worked with 5+ years Rails apps have seen some shit.
27. Maintainable web apps Cargo
28. Cult? Service-Oriented Architecture
29. (I happen to agree with them but that’s not the point.)
30. So, should we just stop using Rails?
31. It’s not quite that simple.
32. I would still use Rails.
33. Mostly because I have no choice.
34. It’s (close to) impossible to find Ruby devs who don’t do Rails.
35. In many cases, Rails is all of the Ruby developers know.
36. Rails’ toolset has become the toolset of Ruby developers.
37. This talk is actually all about Tools.
38. (Sorry for the link bait.)
39. (#sorrynotsorry)
40. Context.
41. In preparing this talk, 3 things collided in my head.
42. 1.
43. I got a new job.
45. Architecture Grand Master High Guild Wizard Poobah* * or some combination thereof.
46. Envato has 75 devs.
47. My job is improving: - Software architecture - Software modeling - Software design
48. Rails’ tools. MVC.
49. ActiveRecord Must die.
50. ActionView Too weak an interface.
51. ActionController Roda is (already) better.
52. ActionMailer Use literally anything else.
53. ActiveSupport Monkey-patches everything.
54. Turbolinks Abomination.
55. Concerns Congratulations! You just made it easier to violate SRP.
56. Asset Pipeline “Terrible.” - Tim Lucas
57. Model, View, Controller. Ranges from to .
58. Envato has 75 Rails devs.
60. Actually, that’s not quite right.
61. We have a bit of everything in production.
62. Rails, non-Rails Ruby, Node.js, Elixir, Go, MySQL, PostgreSQL, MongoDB, Redis, Memcache, ElasticSearch, Sphinx.
63. I cannot control this language proliferation. Nor do I intend to try.
64. Still, my job is improving: - Software architecture - Software modeling - Software design
65. How do software developers get better?
66. Be a better software developer Cargo
67. 2.
68. I talked to someone who wouldn’t apply for a job.
69. Why? No Ruby experience.
70. Lots of solid OO experience.
71. Do I actually care? No.
72. Rails is a hammer.
73. If the person has other tools, they are still very valuable.
74. In fact, not knowing Rails can be a benefit.
75. Diversity of experiences and opinions lead to better outcomes.
76. Be a better software developer Cargo
77. Cult Mastery of a single language and/or framework
78. 3.
84. I fundamentally disagree. (But I wasn’t looking for a shitfight.)
85. is neither: new, nor shiny, nor hipster. New Shiny Hipster Technology
86. It might seem that way, but most are reinventions of the wheel.
87. Examples!
88. Node.js
89. NoSQL
90. React
91. Elixir/Go/Swift
92. (Our Rails “hacks” aren’t new either.)
93. But perhaps more importantly…
94. Elixir is a hammer.
95. Go is a hammer.
97. React is a… ?
99. Seriously, though, who would have thought it would be the JS community who would get excited about immutability?
100. But I digress…
101. Be a better software developer Cargo
102. Cult Learn new languages and/or frameworks
103. Now, this feels like a catch-22.
104. Be a better software developer Cargo
105. Cult Mastery of a single language and/or framework
106. Cult Learn new languages and/or frameworks
107. The thing with languages…
108. …it’s hard to use more than one at a time.
109. Elixir won’t do you much good if I need to drop you into a Rails monolith.
110. Kind of like hammers. They might be somewhat different, but you typically only need one.
111. If all you have is hammers in your tool belt, then you are of no use to me.
112. Diversify your toolset.
113. No tool, no matter how great, is right for every occasion.
114. Accumulate more, and learn when to apply each one.
116. Learn patterns that translate.
117. Patterns are language agnostic and can be applied in different situations with different tech.
118. Languages/frameworks can be be a way to learn patterns.
119. But you don’t have to.
120. Learn new languages and frameworks because of the patterns they employ. (Not because they’re the New Shiny.)
121. Be explicit about what those patterns are.
122. If you can't extrapolate the pattern(s), it's not a useful tool.
123. Because you can’t use it independently to the language/ framework you learnt it from.
124. (Now, I have a list.)
125. Rather than New Shiny Hipster Tech, I want you to learn about: I’ll tweet out a link to all of them later.
126. OLID
127. (SOLID without the S.)
128. (Not that S isn’t important.)
129. OLID: Open/closed principle.
130. Poor fella is very misunderstood.
131. Can’t modify existing code, only extend it.
132. Code that doesn’t have to be changed every time our requirements change.
133. OLID: Liskov substitution principle.
134. OLID: Interface segregation.
135. Small, isolated interfaces.
136. High cohesion/low coupling.
137. (Law of Demeter.)
138. (Postel’s Law.)
139. OLID: Dependency inversion.
140. Depend on things that are of the same, or higher level, of abstraction.
141. Inversion != Injection.
142. Dependency injection is about how to acquire dependencies.
143. Dependency inversion is about what should depend on what.
144. (By all means, learn about Dependency Injection as well.)
145. Separation of Concerns.
146. Not the same as ActiveSupport::Concern.
147. Modularity via encapsulation.
148. Interface design.
149. Event sourcing.
150. Don’t store current state of an object. Store events that have happened to it.
151. Derive an object’s current state by replaying its events.
152. (I’m talking about this at RubyConf.)
153. CQS. CQRS.
154. Command Query (Responsiblity) Segregation.
155. CQS: A method can either mutate state or return state. Never both.
156. CQRS: A class can either mutate state or return state. Never both.
157. Immutable data structures.
158. Gateway to concurrency.
159. Now, immutable data structures in the application layer is nice and all…
160. …but not as awesome as an immutable, append-only persistence layer.
161. (You can do this with any database, don’t have to hipster around with Datomic.)
162. Functional vs. Object-oriented.
163. Only exception that I’ll make.
164. But you don’t actually have to learn a new language to do functional programming.
165. Functional Ruby. It’s a thing.
166. Service-Oriented Architecture.
167. Ewww, Microservices.
168. Not always, but sometimes. Form your own opinion.
169. Asynchronous communication by default.
170. Pseudo-synchronous.
171. Message-oriented middleware (message buses).
172. Idempotent operations.
173. Exactly once. At most once or At least once.
174. Do stuff with data.
175. Data warehousing. Star schemas.
176. MapReduce.
177. And, you know, the actual “patterns.”
178. Read the “Gang of Four” book. (I know it’s old.)
179. Reactor pattern.
180. Strategy pattern.
181. Memento pattern.
182. Factory pattern.
183. Adapter pattern.
184. Observer pattern.
185. At this point I’m just throwing words at you.
186. My point is: all of these are language agnostic.
187. They can be just as fun and interesting as the New Shiny Hipster Tech.
188. And they’re more useful.
189. Learning and understanding these will allow you to consider more than one solution to a problem.
190. If you can only think of one way to solve a problem, the only way you can succeed is by luck.
191. Luck is not a strategy.
192. If you have options, you increase your chance of success.
193. Be a better software developer Cargo
194. (Probably) Cult Evolve language/framework agnostic skills.
195. Languages, frameworks, and all the shiny hipster tech will come and go.
196. Wouldn’t you rather bet on adding more generic software development tools?
197. I would, and I am.
198. Thank you. Questions? Please follow me on the Twitter: @vonconrad