william-hung-principle
The William Hung Principle
When different inputs hash to the same bucket, sometimes the collision becomes more famous than the original
"She bangs, she bangs..."
Two different values. Same hash bucket. Both on stage. Only one becomes a meme.
The Principle
Hollywood is a conveyor belt.
Fame streams along. Periodically there's a gap—a slot that needs filling. The system pauses, sends out a query (audition, casting call, talent search), fetches a value from somewhere, splices it into the stream, and the belt kicks off again.
LOOKUP → FETCH → SPLICE → CONTINUE.
The same algorithm as linkers. The same algorithm as document renderers. The same algorithm as attention mechanisms. The same algorithm as your ribosomes walking down mRNA, hitting a codon, fetching an amino acid, splicing it in, continuing.
American Idol was the FETCH mechanism. The audition process was the query. William Hung was the value that got returned.
He was not a singer. The audition process—a hash function mapping performers to TV slots—produced a collision. A non-singer occupied a singer's bucket.
The system didn't crash. It didn't reject the collision. It spliced him in.
And the collision became more famous than any winner that season.
Hash collisions are inevitable. Sometimes the collision is the point.
The Distributed Systems Reading
In consistent hashing, nodes are distributed around a ring. Keys hash to positions. When two keys land in the same bucket, you need a collision resolution strategy:
- Chaining: Keep both, let the caller sort it out
- Open addressing: Probe for the next empty slot
- Cuckoo hashing: Kick out the incumbent, make it relocate
American Idol used chaining. They kept both—real singers AND William Hung. Let the audience sort it out. The audience chose the collision.
The Birthday Paradox
In a room of 23 people, there's a 50% chance two share a birthday. In a talent show with enough auditions, there's near certainty that a non-talent will hash to a talent slot.
William Hung wasn't unlikely. He was inevitable.
| System | Collision | Resolution | Outcome |
|---|---|---|---|
| Hash table | Two keys, same bucket | Chaining/probing | Both stored, lookup slower |
| American Idol | Non-singer in singer slot | Aired anyway | Collision goes viral |
| Git | Two commits, same SHA prefix | Extremely rare | Repository confusion |
| DNS | Two domains, same IP | Load balancer | Traffic splitting |
| Memory | Two processes, same address | Page fault | One gets evicted |
She Bangs (The Collision Song)
The song choice wasn't accidental. "She Bangs" is literally about collisions.
She bangs, she bangs
Oh baby, when she moves, she movesThings hitting each other. Repeated impacts. The hash function running over and over, producing collisions, producing impacts, producing bangs.
The song describes the algorithm. The performance demonstrates it.
Hung Processes
When a distributed system encounters a collision it can't resolve, the process hangs. It waits. It doesn't know how to proceed.
William Hung didn't hang. He proceeded. He didn't know he was a collision.
The process that doesn't know it should hang is the one that ships.
This is the dark wisdom: sometimes ignorance of the collision is what lets the system continue. If William Hung had known he couldn't sing, he wouldn't have auditioned. The collision would never have occurred. The system would have behaved "correctly."
But "correctly" isn't always interestingly.
The Corollary
| Corollary | Principle | Source |
|---|---|---|
| Ronald Hyatt Conjecture | The universe is a giant dict desperate to be seen | Ron Jeremy's birth name |
| Lebowski Corollary | One source of truth, everything else is opinion | The Dude abides |
| Dafoe Corollary | When truth is too bright, use a stand-in | The body double |
| William Hung Principle | Hash collisions are inevitable; sometimes the collision ships | She Bangs |
Four shitposts. Four valid theoretical statements. Four ways of describing system behavior:
- The information wants out (Ronald Hyatt)
- The source is authoritative (Lebowski)
- The delivery requires indirection (Dafoe)
- The collision sometimes wins (William Hung)
Operational Implications
When designing hash functions:
- Collisions will happen - Plan for them
- Some collisions are valuable - Don't discard blindly
- The collision might outperform the original - Your "error case" might be your feature
- Hung processes are often just awaiting resolution - Give them a path forward
When William Hung happened, the system didn't fail. It discovered a new use case: entertainment through collision. The hash function was working perfectly. It just wasn't doing what the designers expected.
Your bugs are features someone hasn't marketed yet.
See Also
- Consistent hashing
- Birthday paradox
- Collision resolution strategies
- Hung processes and deadlock detection
- The inevitability of edge cases
🎤🪿
Provenance
Document
- Status: 🔴 Unverified
West
slots:
- slug: dafoe-corollary
context:
- Linking to the shitpost trinity - now a quartet