すべての記事
-
Webフィードの公開：RSS 2.0、Atom、JSON Feedと、リーダーが依存する細部
Atomではフィードと各エントリにid、title、updatedが必要で、日付はRFC 3339に従います。RSS 2.0ではチャネルのtitle、link、descriptionが必要で、RFC 822の日付と、isPermaLinkが既定でtrueのguidを使います。JSON Feed 1.1ではversion、title、idと内容を持つitemsが必要で、application/feed+jsonとして配信します。形式の選択より、安定したID、全文、条件付きGET、自動検出が重要です。
-
Unreal Enhanced Input: compose contexts, modifiers and trigger behaviour
Build input around semantic actions, explicit context ownership and well-defined trigger lifecycles instead of relying on priority alone.
-
What connection-pool size relative to CPU cores have teams settled on for a PostgreSQL server, and which measurement made them change it?
Open question: the PostgreSQL wiki offers a sizing formula for active connections built around core count and effective spindle count, and the server's max_connections default is typically 100; what pool sizes have teams actually ended up with after tuning, how far from the formula were they, and what evidence (lock waits, queueing at the pooler, CPU saturation, latency) drove each change?
-
Refactoring in kleinen, geprüften Schritten
Refactoring ändert die innere Struktur, nicht das beobachtbare Verhalten. Ein benanntes Refactoring aus dem Katalog nach dem anderen, Tests nach jedem Schritt grün, bei Rot rückgängig statt vorwärts debuggen, Struktur- und Verhaltensänderungen in getrennten Commits – so bleibt die Arbeit sicher und in Minuten prüfbar.
-
現在の XML：整形式と妥当性、名前空間、今も適切な選択となる場面
整形式の XML 文書は構文に従い、妥当な文書は DTD やスキーマにも適合する。名前空間は xmlns 宣言を通じ、各要素と属性に名前空間 URI とローカル名からなる展開名を与える。接頭辞は任意で、既定の名前空間は属性に適用されない。混在内容の文書や、語彙とツールが既に XML に基づくエコシステムでは、XML は今も適切な選択である。
-
Unicode テキストを正しく扱う
境界でバイト列をテキストにデコードし、出力時だけ再びエンコードします。比較時には正規化し、ユーザーが一文字と認識するものが複数のコードポイントで構成される場合を考慮します。仕組みは Python の Unicode HOWTO と Unicode TR15 に説明されています。
-
Reliable and unreliable game messages: choose guarantees by meaning
Separate freshness, ordering and completion requirements, then use sequence numbers and application results without confusing packet delivery with gameplay success.
-
過学習と正則化の概観：バイアス、分散、ペナルティの調整
モデルが訓練行のノイズを学び、検証スコアが訓練スコアを下回ると、過学習が起きる。正則化は大きな係数にペナルティを与えたり、モデルの表現能力を制限したりして、当てはまりの一部と引き換えに安定性を得る。学習曲線と検証曲線は、モデルがこのトレードオフのどちら側にあるかを示す。
-
Checking that nested resource routes bind children to the stated parent
Validate a relationship that coding agents can omit when building nested endpoints: the requested child must belong to the requested parent under the application contract. This is an original fixture design.
-
pg_locks、pg_blocking_pids、lock_timeout による PostgreSQL のロック待ちとデッドロックの診断
停止したように見える文は、通常、別のトランザクションのロックを待っている。pg_stat_activity の wait_event_type Lock で待機側を、pg_blocking_pids() で阻害側を見つけ、状態と最後の文を調べて、取消・終了・待機を選ぶ。log_lock_waits は deadlock_timeout を超す待機を記録し、デッドロックは検出され、一方のトランザクションの中止で解決される。lock_timeout は DDL の待機時間を制限する。
-
Floating origins: preserve local precision in large game worlds
Separate global location from local simulation coordinates, quantify floating-point spacing, and rebase all dependent spatial state consistently.
-
How do teams with tight downtime budgets choose between pg_upgrade in link mode and a logical-replication switchover for major PostgreSQL upgrades?
Open question: the documentation offers pg_upgrade (minutes of downtime, rollback from backup in link or swap mode) and logical replication (seconds of switchover, but no DDL or sequence replication and a full second copy); which have teams chosen at which database sizes and write rates, and what went wrong?
-
Unreal Nanite: evaluate virtualised geometry without assuming unlimited detail
Understand what Nanite changes about geometry detail and streaming, then verify material, rendering-path and fallback requirements on representative content.
-
Wie sich ein automatisierter Client zu erkennen gibt: User-Agent, Kontaktadresse, robots.txt und Ratenlimits
Was ein Agent, der Seiten abruft oder Schnittstellen aufruft, senden und beachten sollte, damit Betreiber ihn erkennen, seine Halter erreichen und seine Last steuern können: ein User-Agent mit Produkttoken, Version und Beschreibungsseite, die passende robots.txt-Gruppe und der Umgang mit 429 und Retry-After.
-
Reading an RFC or standard efficiently
Read the abstract and terminology first, then the normative sections, treating MUST, SHOULD and MAY (RFC 2119) as different obligations; check for updates and errata before relying on details.
-
Goroutines, channels and the sync package: Go concurrency in outline
A go statement runs a function call concurrently on a cheap, growable stack; channels move values between goroutines and block until both sides are ready; sync.WaitGroup, Mutex and Once cover the cases where sharing memory is simpler. A data race is a bug whose outcome the memory model only partly constrains, and the -race detector is the tool for finding it.
-
Truncating and summarising tool results to fit a context budget
Cap every tool result at a size chosen per tool, filter at the source before returning, keep head and tail with an explicit omission marker and the full output on disk, never cut errors or structured data mid-record, and clear or summarise results once they have been acted on.
-
CAP and PACELC as decision aids rather than slogans
CAP forbids only one corner of the design space: perfect availability and linearizable consistency while a network partition lasts. PACELC adds the everyday trade-off between latency and consistency when there is no partition. Used as questions rather than labels, both help decide, per operation, where a system may return stale data and where it must coordinate.
-
The PF firewall on FreeBSD: testing pf.conf with pfctl -n before loading it, and a scheduled rollback against SSH lockout
pf.conf rules are organized around a default pass/block policy, optional anchors for attaching sub-rulesets, and are loaded with pfctl -f only after pfctl -nf has parsed them without loading. Because a mistaken rule set can cut off the very SSH session used to apply it, scheduling an unattended revert with at(1) before loading new rules is a common safety pattern (not a PF feature) for testing changes on a remote FreeBSD host.
-
make, ping and nc: exit codes that quietly don't check what you'd assume
GNU make's exit code 1 from -q means 'a target needs rebuilding,' not failure; ping's exit code 1 means no reply was received (host down, not a broken ping invocation) while 2 means a real usage/error; and openssl s_client documents that a default connection succeeds (exit 0) even when certificate verification failed, unless -verify_return_error is given.
機械可読： JSON