Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Limits of Modularity
Search
Samuel E. Giddins
May 19, 2015
Technology
0
77
Limits of Modularity
Samuel E. Giddins
May 19, 2015
Tweet
Share
More Decks by Samuel E. Giddins
See All by Samuel E. Giddins
The Challenges of Building a Sigstore Client from Scratch
segiddins
0
55
Keeping the Gems Sparkling
segiddins
0
40
A Survey of RubyGems CVEs
segiddins
0
40
Handling 225k requests per second to RubyGems.org
segiddins
0
71
State of the RubyGems 2023
segiddins
0
100
Building Broken Gems
segiddins
0
74
Switching Disciplines as a Tech Lead
segiddins
0
40
Source Code to Executable
segiddins
0
81
Empowering iOS Developers
segiddins
1
85
Other Decks in Technology
See All in Technology
Snowflake Summit 2025全体振り返り / Snowflake Summit 2025 Overall Review
mtpooh
2
420
解析の定理証明実践@Lean 4
dec9ue
1
190
Yamla: Rustでつくるリアルタイム性を追求した機械学習基盤 / Yamla: A Rust-Based Machine Learning Platform Pursuing Real-Time Capabilities
lycorptech_jp
PRO
4
160
生まれ変わった AWS Security Hub (Preview) を紹介 #reInforce_osaka / reInforce New Security Hub
masahirokawahara
0
320
Prox Industries株式会社 会社紹介資料
proxindustries
0
340
Wasm元年
askua
0
160
Github Copilot エージェントモードで試してみた
ochtum
0
110
mrubyと micro-ROSが繋ぐロボットの世界
kishima
2
370
"サービスチーム" での技術選定 / Making Technology Decisions for the Service Team
kaminashi
1
200
Oracle Audit Vault and Database Firewall 20 概要
oracle4engineer
PRO
3
1.7k
AI導入の理想と現実~コストと浸透〜
oprstchn
0
120
20250625 Snowflake Summit 2025活用事例 レポート / Nowcast Snowflake Summit 2025 Case Study Report
kkuv
1
340
Featured
See All Featured
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
138
34k
Put a Button on it: Removing Barriers to Going Fast.
kastner
60
3.9k
What’s in a name? Adding method to the madness
productmarketing
PRO
23
3.5k
A Tale of Four Properties
chriscoyier
160
23k
Visualization
eitanlees
146
16k
Optimising Largest Contentful Paint
csswizardry
37
3.3k
VelocityConf: Rendering Performance Case Studies
addyosmani
331
24k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
34
5.9k
KATA
mclloyd
30
14k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
107
19k
Measuring & Analyzing Core Web Vitals
bluesmoon
7
490
A Modern Web Designer's Workflow
chriscoyier
694
190k
Transcript
The Limits of Modularity Samuel Giddins
Samuel Giddins Realm CocoaPods Bundler RestKit
Bundler & CocoaPods Two Dependency Managers
None
• ruby -e "$(curl -fsSL https:// raw.githubusercontent.com/Homebrew/install/master/ install)" • brew
install python • easy_install pip • pip install pygments • brew install rbenv ruby-install • rbenv install 2.2.2 • gem install pygments.rb
Let’s take a peek at the architecture behind CocoaPods
None
In an iOS App
Let’s start at the networking layer
We have ’s NSURLConnection / NSURLSession to base things off
of So far, so good
We then use JXHTTP or AFNetworking, since both provide some
features and a nicer API
We then make a TMAPIClient (inside a pod called TMTumblrSDK),
since we want to abstract away the network calls we make
We then make a TumblrCore pod, to be able to
share code between our app and its extensions
And now, we have Tumblr.app! Whew.
That’s what’s called a deep dependency graph.
aside: Those things are really hard to fully ‘resolve’ It’s
a problem I’ve sunk many hours into
So, this is great! We’ve re-used lots of code, made
things modular, ‘componentized’ everything... Life is wonderful!
Until...
We need to change things. (And yes, this is where
we get to the interesting parts)
Let’s say we want to update our UI to show
upload progress for a new post
(I went through nearly this exact scenario when I worked
on the Tumblr app)
OK, so the first step is going to be figuring
out where, exactly, in this stack we can even start.
Start in the wrong place, and you’ll be working at
the wrong level of abstraction, and you’ll be contorting to even make simple changes.
I just need to rearchitect sockets and everything will work
beautifully!
Have fun.
Well, I just need to swizzle these three methods, and
I can do everything in my app!
And again in my share extension. And then again in
the right place because who in their right mind wants to swizzle the same method in three places!?
Again & Again & Again
So we settle upon making this change inside JXHTTP. Excellent.
@property JXHTTPRequestProgressBlock progressBlock;
While we're in there, we choose to move a few
things around to make adding support for that progressBlock easier.
And this is great! !
So now we release a new minor version, and go
back up one level to incorporate those changes.
s/~> 1.2.4/~> 1.3.0
And our tests now fail !
Imagine this happening all the way up the dependency chain
Dependency Paralysis
When our frameworks aren't perfect, the gaps between them can
easily fissure
And our frameworks are never perfect
Edge cases and implicit assumptions propogate through each dependency
One subtle change can be in just the wrong place
And everything collapses
Our usual measures for keeping things sane & safe will
work against us
• Tests • Documentation • Encapsulation • Stable releases
All of those work by applying the brakes on drastic
changes
But N levels deep, every change at the bottom is
drastic
And you'll find that almost every change requires touching something
near the bottom
We also have to decide where to put everything
And choosing wrong? => !
We hear stories of Facebooks and the like with their
mono-repos
And all of this is why Move fast vs. being
"really" modular
So, Samuel, what's the right answer?
Available now on Speaker Deck. https://speakerdeck.com/segiddins/limits-of-modularity
for (Question *question in self.audience.questions) { NSLog(@"%@", [[question ask] valueForKey:@"answer"]);
}
Samuel Giddins Realm @segiddins