Kerning
Adjusting the space between specific pairs of letters so that text appears evenly spaced -- distinct from tracking, which adjusts spacing uniformly -- encoded in fonts through kern tables and OpenType's GPOS positioning data.
Plain-language explanations of the ideas behind software - compilers, garbage collection, objects, types.
Adjusting the space between specific pairs of letters so that text appears evenly spaced -- distinct from tracking, which adjusts spacing uniformly -- encoded in fonts through kern tables and OpenType's GPOS positioning data.
The problem and protocols for two parties to establish a shared secret key securely, the classic obstacle for symmetric encryption that Diffie-Hellman solved.
The simplest NoSQL model: a distributed hash map of unique keys to opaque values, optimized for fast, scalable lookups.
Hacker term, also spelled kluge, for an inelegant solution that nonetheless works: a clumsy patch or clever programming trick that handles a nasty case efficiently if not cleanly, central to the engineering folklore collected in the Jargon File.
An anonymous function written inline without a name, such as x => x + 1; named after the lambda calculus and now found in nearly every modern language.
Church's formal system of function definition and application that is a universal model of computation and the theoretical basis of functional programming.
Delaying the evaluation of an expression until its value is actually needed, which enables infinite data structures and can avoid unnecessary work.
The process by which distributed nodes choose a single node to coordinate, and choose a new one when it fails.
The most common replication scheme: one leader accepts writes and streams them to followers that serve reads.
To keep clocks aligned with the Earth's uneven rotation, an extra second (shown as 23:59:60) is occasionally inserted into UTC. Software that assumes every minute has exactly 60 seconds can break when it sees the 61st, and the disruption pushed metrology bodies to vote in 2022 to phase the leap second out by 2035.
The simplest search: check each element in turn until the target is found or the data is exhausted, running in O(n) time.
The strongest single-object consistency condition: every operation appears to take effect instantaneously at one moment between its call and its return, so the system behaves like a single up-to-date copy.
A sequence of nodes each pointing to the next, giving cheap insertion and deletion at a known position but slow access by index.
A Linux kernel feature that gives a process group its own isolated view of system resources such as process IDs, network stacks, mount points, and user IDs.
The empirical tendency of programs to reuse data they accessed recently (temporal locality) and to access data near what they just touched (spatial locality) - the regularity that makes caches and the whole memory hierarchy work, formalized by Peter Denning's 1968 working-set model.
A generated file that pins the exact resolved versions of every dependency so installs are reproducible across machines and over time.
A counter-based mechanism introduced by Leslie Lamport in 1978 for ordering events across a distributed system without relying on synchronized physical clocks.
The binary instructions a CPU fetches and executes directly, encoded as opcodes and operands; the layer beneath assembly language that every higher-level program is ultimately turned into.
The practice of writing programs that generate or transform other programs, exemplified by Lisp-family macros that operate on code as data to extend the language itself.
Code whose execution is overseen by a runtime such as the CLR or the JVM, which handles memory, type safety, and security - in contrast to native or unmanaged code compiled directly to machine instructions.
Three higher-order functions over collections: map transforms each element, filter keeps elements passing a test, and reduce combines elements into a single result.
An optimization that caches the result of a function call so that later calls with the same arguments return the stored answer instead of recomputing it.
The layered arrangement of storage in a computer - registers, cache, main memory, disk - trading speed and cost against capacity, so that fast memory holds the data in active use and slow memory holds the rest.
A property of programs that prevents bugs like use-after-free, buffer overflows, and data races, achieved through approaches such as garbage collection or compile-time ownership checks.
A stable, divide-and-conquer sorting algorithm that splits a list, sorts the halves, and merges them, with a guaranteed O(n log n) running time.
Middleware that lets systems communicate asynchronously by placing messages on a queue for other systems to consume, decoupling senders from receivers.
Coordinating concurrent processes by sending each other messages instead of sharing memory, eliminating locks and data races by design.
An internal layer of micro-instructions inside a CPU that implements its complex machine instructions in terms of simple hardware steps; proposed by Maurice Wilkes in 1951 and now used to patch processor errata in the field.
The cheapest cycle-free set of edges that connects every node of a weighted graph, found by greedy algorithms from Kruskal (1956) and Prim (1957).
The practice of applying DevOps discipline to machine learning - versioning data and models, building CI/CD pipelines for ML, and monitoring deployed models for drift.
A stand-in for a real dependency in a test, programmed with expectations about how it should be called, so a unit can be tested in isolation from slow or unpredictable collaborators like databases and networks.
The practice of deploying a trained machine learning model behind an API for inference, spanning batch and online serving and confronting the training-serving skew problem.
A functional design pattern, borrowed from category theory, for sequencing computations that carry a context such as state, failure, or input/output while keeping functions pure.
Keeping many projects or services in a single version-control repository, as Google and others do, trading tooling complexity for unified change and dependency management.
Gordon Moore's 1965 observation that the number of components on an integrated circuit doubles at a steady, predictable rate, which became the planning roadmap of the semiconductor industry.
Requiring two or more independent proofs of identity from different categories - something you know, have, or are - so that a stolen password alone is not enough to log in.
A single running software instance or pool of infrastructure serving many customers ('tenants') at once, with each tenant's data isolated from the others - the economic engine that makes cloud and SaaS cheap.
Keeping multiple versions of rows so readers see a consistent snapshot without blocking writers, and writers do not block readers - how PostgreSQL and Oracle achieve high concurrency.
The per-project folder where npm installs a package's full dependency tree, famously huge, and the reason flat installs, hoisting, and pnpm's linking exist.
Organizing tables so each fact is stored once, by decomposing them according to their functional dependencies into successive normal forms.
The tendency of a team or organization to reject external solutions and rebuild them in-house, out of pride, distrust, or insularity.
The class of the hardest problems in NP, such that a fast algorithm for any one of them would yield a fast algorithm for every problem in NP.
Code that is deliberately made hard to read, whether as an art form and competitive sport, as a way to protect intellectual property and hide malware, or as a side effect of compressing source down to nothing.
A way of organizing programs around objects that bundle data together with the behavior that acts on it, with roots in Simula and Smalltalk.
A layer that maps database tables to objects in code so developers work with objects instead of writing SQL by hand.
A behavioral design pattern in which a subject keeps a list of dependent observers and notifies them automatically when its state changes.
The split between transaction processing (many small reads and writes) and analytical processing (large aggregate queries over history), two workloads so different they use different database designs.
Encryption with a truly random key as long as the message and used only once; the only cipher proven to offer perfect secrecy, but impractical because of the key-distribution burden.