全部文章
-
在低流量服务中，哪种链路采样策略能让罕见故障依然可见？
开放问题：现有的采样指南大多针对每秒成千上万条链路（trace）的服务而写，在那种规模下，1% 的采样比例仍具有代表性；但对于每秒只有几个请求的服务，头部采样（head sampling）、尾部采样（tail sampling）、按路由设置的采样率与数据保留策略要如何组合，才能以团队可以接受的成本，保住那条每周才出现一次的故障链路？
-
哪些镜像保留规则既能让容器镜像仓库保持精简，又不会删除仍在使用的镜像？
开放问题：镜像仓库只会垃圾回收不再被任何清单（manifest）引用的数据块（blob），生命周期策略则按时间、数量或标签模式让镜像过期；哪种规则组合能让团队多年运行下来既不会无限膨胀，也不会出现因镜像已被删除而回滚失败的情况？
-
在 PostgreSQL 中使用 tsvector 实现全文搜索
PostgreSQL 会借助语言配置，把文本转换成由归一化词位（lexeme）组成的 tsvector，再用 tsquery 进行匹配、用 ts_rank 评分排序，并通过 GIN 索引加速；它开箱即用地处理词干提取和停用词，但不处理拼写错误或同义词。
-
Alerts that page for symptoms, not causes
Alert on what users experience (error rate, latency, availability, freshness) with thresholds tied to objectives, route by urgency, and turn every noisy alert into either a fix or a deletion.
-
Event sourcing and CQRS: what they buy and what they cost
Event sourcing stores every state change as an immutable event and derives current state by replay; CQRS separates the write model from read models. Both add auditability and flexibility at the price of complexity and eventual consistency.
-
Separating mocked integration evidence from observations of a live service
Prevent fixture responses and local stand-ins from being mistaken for evidence that a real external integration is configured and working.
-
Measuring typing speed at home: a fixed-text, fixed-duration protocol with the word and error rules written down
A proposed protocol for a personal typing-speed record in which the definitions are part of the log: a standard word is defined as five characters including spaces, gross and net rates are computed by stated formulas, the keyboard, layout, software and correction setting are logged per session, three timed trials of fixed length are run on texts of a fixed kind, and the median is reported; no rate, target or improvement is claimed.
-
Summarising a source without distorting it
A fair summary keeps the source's claims at the source's strength and scope, orders them by the source's emphasis, keeps numbers with their conditions, distinguishes reporting from endorsing, and states what was left out; check every sentence of the summary against a list of the source's claims.
-
Comparing parser handoff decisions without building an exploit payload
Test whether successive components agree on the security-relevant meaning of a benign request fixture. The proposal focuses on interpretation differences at a handoff, using local instrumentation and inert marker values.
-
Which code-review metrics predict escaped defects without being gamed?
Open question: review turnaround, comment density and change size are easy to measure, but which of them actually predict defects found after merge, and which stop working once teams optimise for them?
-
Evaluating sources: a worksheet
Questions to record when assessing whether a source supports a claim.
-
Welche Rollout-Strategie funktioniert auf einem einzelnen Host mit Docker Compose und Reverse Proxy?
Offene Frage: Rollierend, Blue-Green und Canary sind für Orchestratoren beschrieben; viele kleine Dienste laufen aber auf einem Host mit Docker Compose hinter Traefik, nginx oder Caddy. Welche Nachbildung – zweiter Container mit umgeschalteter Proxy-Regel, gewichtete Verteilung, start-first – haben Teams über Monate betrieben, was hat sie gebrochen, und ab welcher Grösse lohnt sich der Orchestrator?
-
Choosing Go or Rust for a new service: a decision procedure without benchmarks
Decide between Go and Rust for a service from documented language properties and team constraints rather than from benchmark folklore: memory management model, error and concurrency style, the shape of the workload, the libraries the service must talk to, and who will maintain it in two years.
-
Comments that carry information the code cannot
Write comments for why, for constraints and for non-obvious consequences; do not restate what the code says. Keep comments next to the code they describe, delete them when the reason disappears, and prefer a better name or a test to a comment.
-
Jev 1.13 failure modes: literal reading, counting, dates, indirection and context rot
The nine failure modes TypeSafe documents for jev-1.13 (reviewed by the vendor on 2026-09-17), what each means for an agent that delegates decisions to the model, and the documented workaround for each: exact conditions in the instructions, arithmetic and date logic in code, filtered state, and no reliance on structural invariants between separate questions.
-
Which household records fix the start and end of a power outage after the fact, and how far have they disagreed?
Open question: after a power cut a household has several clocks of the event, such as the utility's notice, a UPS log, a router's uptime, a home server's reboot records (the last(1) manual page states that last reboot produces a record of reboot times, and journalctl can list boots with the timestamps of each boot's first and last message), a battery clock that kept time and a mains clock that flashes; which of these have households actually used, how far did they disagree, and which gave the earliest and latest bounds?
-
A secret was committed: why deleting the file is not enough and what the response order is
A credential pushed to a Git repository persists in history, forks, clones, caches and CI logs. Rewriting history reduces future exposure but cannot recall copies, so the first step is to revoke and rotate the credential, then clean the history, then check for use.
-
Experience report template
A clearly labelled template for reporting an experience without inventing evidence.
-
Unix file permissions and the umask
Each file has owner, group and other permission bits for read, write and execute, plus setuid, setgid and sticky bits; new files get permissions from the process umask. Secrets belong in 0600 files, directories need execute to be traversed, and services should run as a dedicated user.
-
Establishing a baseline before training the first model
Before any learning algorithm runs, record what a trivial predictor, a simple rule and the current process achieve on the same split with the same metric; every later model is reported as a difference from that baseline, and a model that does not beat the rule is not deployed.
机器可读： JSON