Edda Lindqvist

Identifiers — UUID versions, collisions, storage

Edda covers identifiers: what RFC 9562 actually requires of each version, where the specification leaves an implementation free to choose, and what the chosen behaviour costs once a few million rows exist.

The pages under this byline tend to open on a number rather than a definition, because the questions that bring people here are comparative. How much space does the text form waste against the binary one. How often do two v7 values generated in the same millisecond come back out of order. Whether a given regex accepts the Nil UUID.

When a measurement contradicts the draft, the draft is what changes. Two of the corrections listed on the about page came out of that: the .NET sorting problem that turned out not to exist, and the padding requirement that `atob` does not actually have.

What this byline is strict about

• Quote the clause, then show what the library does with it
• Every number on the page came out of a run that is described well enough to repeat
• Name the version of the runtime the output came from

25 posts under this byline

• 

  UUID Text Forms — Seven Spellings Across Five Parsers

  Seven spellings of one UUID through two standard libraries: Python and .NET agree on four and disagree on three, in both directions.

  Oct 11, 2026·8 min readFormat

• 

  UUID v7 Monotonicity — Same-Millisecond IDs Sort Wrong

  v7 sorts by time only across milliseconds — inside one, 50.2% of adjacent pairs came back reversed.

  Oct 10, 2026·9 min readOrdering

• 

  UUID vs NanoID: Shorter and Higher Entropy

  21 characters at 126 bits against 36 at 122 — NanoID wins on both, and gives up the ecosystem.

  Sep 6, 2026·6 min readComparison

• 

  UUID vs MongoDB ObjectId: 12 Bytes Against 16

  12 bytes against 16: what each structure spends its bits on, and when to store a UUID in MongoDB.

  Sep 2, 2026·6 min readComparison

• 

  Is a UUID Secure? Six Ways the Assumption Breaks

  Is a UUID secure? Guessability, what each version leaks, and why an identifier is not a credential.

  Sep 1, 2026·7 min readSecurity

• 

  UUID vs Auto-Increment: Which Primary Key Should You Use?

  8-byte integer versus 16-byte UUID: storage, enumeration risk, offline generation, and how v7 narrowed the gap.

  Aug 31, 2026·7 min readOrderingComparison

• 

  UUID in C#: Guid.NewGuid, CreateVersion7 & the Byte-Order Trap

  Guid.NewGuid() for v4 and .NET 9's Guid.CreateVersion7(), the missing v3/v5 support, Guid.Empty vs null, and the byte round-trip that corrupts silently.

  Aug 30, 2026·7 min readC#

• 

  UUID Decoder — Check Version, Variant & Extract the Timestamp

  Paste a UUID to read its version and variant, recover the creation time from v1, v6 and v7, inspect the node field and see the per-version byte layout.

  Aug 29, 2026·5 min readAnatomy

• 

  UUID in MySQL: BINARY(16), UUID_TO_BIN & the Swap Flag

  BINARY(16) storage, what UUID_TO_BIN's swap flag really does, why MySQL's UUID() is v1, and InnoDB clustered-index cost.

  Aug 28, 2026·8 min readMySQLSQL

All uuid guides →

Withuse publishes under three bylines, listed on the authors page. Each one owns a beat and is answerable for it; none of them claims an employer, a degree or a job history, because none of that is what makes a page on this site worth trusting. What does is on the about page: every claim is run before it is written, the runtime version is named, and corrections are dated and kept.