Computing History · Cluster
Lessons from Failures
Documented projects that failed and why.
- 01 Optimistic Concurrency and the Conflict Check That Arrives Too Late How version-token conflict detection works in a replicated store, why it cannot prevent a conflicting write, and what the abandoned Bayou and Coda designs revealed about the cost of checking after the fact.
- 02 Isolation Levels and the Anomalies They Were Supposed to Prevent Why the SQL standard isolation levels are defined by the anomalies they should exclude, why that definition is incomplete, and what Critic and ARIES work revealed about writes that are visible before they are durable.
- 03 Header Files as an Interface Contract and the ODR Landmine What the ISO C and C++ translation-unit models formalize, how the one-definition rule fails silently when headers disagree, and why the language committee took decades to admit that separate compilation is not composition.
- 04 Source Code Control Before Merging: RCS, CVS and the Locking Model How the first source control systems represented revisions as deltas against a single file, why CVS made concurrent editing possible without solving it, and what the checkout-edit-commit cycle could never guarantee.
- 05 The Floppy Sector Interface as a Lossy Abstraction How a disk controller invented to address one track at a time became the universal disk interface, and why the geometry, translation and sector-numbering decisions made for it still surface in firmware and partition tables.
- 06 X.400 Versus SMTP: When a Better-Designed Messaging System Loses How the OSI message handling system specified store-and-forward routing, naming and delivery reports, why its design was technically more complete than SMTP, and what prevented it from displacing an internet mail architecture that was already deployed.
- 07 The IBM PC BIOS Standard and the Cost of Freezing a Firmware Interface What a ROM-resident interrupt interface guaranteed to software, how its fixed data structures blocked hardware progress, and why the replacement took the form of an emulated decades-old machine rather than a new specification.
- 08 Fitting FDDI Management into an 802 Frame: Ancillary Data and the Unused ALT Field How FDDI carried station management frames, what the alternative media access control field was specified to select, and why a general-purpose slot in a frame header did not produce the architectural flexibility its designers expected.
- 09 Threads, Scheduler Activations and the Failed Promise of User-Level Scheduling How user-level threads multiplex onto kernel threads, what scheduler activations promised in exchange for replacing the kernel scheduler, and why the mechanism was displaced by kernel-managed blocking without ever being implemented widely.
- 10 OAuth 2.0 and the Implicit Grant: Removing a Token from the Browser Why the implicit grant existed, what the browser redirect model made impossible to secure, and how the password and hybrid grants traded a front-channel leak for a different attack surface rather than eliminating it.