How Not to Build a Service
summary
Manage 1.0 targets the wrong market (Enterprise) and costs years to rewind; refocus to Rails users, ditch enterprise-only features like custom proxy and homegrown HTTP encryption, and add integrated online billing instead of a fax checkout. The original client is Java/C at ~70,000 lines; the new Ruby client is ~90% smaller and 40% more memory efficient. The codebase lacks no tests and Ruby expertise, and the product never ate its own dog food, killing trial conversion.
AI-written summary · about
chapters (27)
- 0:00Intro
- 0:28Who am I
- 0:51Rails NV Approved
- 1:10Our Products
- 2:15What Not to Expect
- 2:47Outline
- 4:12Hiring
- 5:15Avoiding Failure
- 5:47Types of Mistakes
- 6:37Business Mistakes
- 7:03Knowing Your Customer
- 8:06Taking Our Own Rails
- 9:23The Trial
- 10:30Social Mistakes
- 11:07No Rails Experience
- 13:01No Testing
- 14:36We Dont Eat Our Own Dog Food
- 16:02Technical Mistakes
- 16:26Client
- 18:44Extra Features
- 19:18Custom Proxy
- 20:30Integrated Billing
- 24:10Groupthink
- 25:33Optimism
- 26:40Recap
- 28:17Lessons Learned
- 29:37Contact Info
related talks
28mNitty Gritty Service Building
22mNitty Gritty Service Building
23mAbstracting a Service - A Case Study in Forward Thinking with SOA
25mI've Made a Huge Mistake: We Did Services All Wrong
33mServices and Rails: The Sh*t They Don't Tell You
36mAn Iterative Approach to Service Oriented Architecture
40mService Architectures for Mere Mortals
40mFacing the Monolith Overcoming Monolithic Applications with SOA