July 9, 2026
Sort drama? Spill the bytes
Common prefix skipping, adaptive sort
Oracle’s old speed boast is back, and commenters are already yelling “prove it”
TLDR: An ex-Oracle engineer says a once-patented data-sorting method that supposedly made Oracle much faster is now free for others to use. But commenters are fixated on one drama point: if Oracle discouraged outside speed testing, how much should anyone trust the old “5x faster” brag?
A long-buried Oracle sorting trick just got a second life after its patent expired, and the inventor is basically saying: someone please build this in open source already. The pitch is juicy even for non-database people: this was a behind-the-scenes way to organize data faster, Oracle said it made things roughly five times quicker, and it was clever enough to start handing over sorted results before it had fully finished. The inventor even wants a snappier name for it than the mouthful of technical labels, floating “Orasort” like a proud parent bringing a baby name to the group chat.
But the real fireworks are in the community reaction, where the applause gets interrupted by a giant, skeptical record scratch. The hottest response is basically: sure, Oracle says it was faster — and who exactly was allowed to check? One commenter went straight for the corporate jugular, pointing out that Oracle licenses have famously discouraged benchmarking, which turns every old performance claim into a trust exercise. That instantly shifts the mood from “wow, cool invention” to “receipts or it didn’t happen.”
There’s also a layer of comedy in the backstory itself: a future Oracle feature born from late-night thinking, hacked together on an old Mac running Linux, then tested across a museum of early-2000s machines. Add in the deliciously shady detail that a crusty old Pentium beat a fancy Sun server, and commenters have all the ingredients they love: nostalgia, corporate side-eye, and a benchmark fight waiting to happen.
Key Points
- •The article says Oracle patent US7680791B2 has expired and describes a sort algorithm the author says was introduced in Oracle 10gR2.
- •The algorithm is described as using common-prefix skipping, adaptive switching between quicksort and MSD radix sort, key-substring caching, and the ability to emit sorted output before the full sort completes.
- •The author frames the work as a response to database sorting needs involving keys larger than 8 bytes and long shared prefixes, which differed from benchmark-focused assumptions.
- •An early proof-of-concept was developed on a PowerPC-based Mac running Yellow Dog Linux and later tested on multiple systems including Pentium III, UltraSPARC IV/IV+, and PA-RISC hardware.
- •The author reports that the new sort outperformed other tested algorithms on larger keys and notes that one implementation issue was that the old Oracle sort was stable while the new one was not.