`rel="next"` from the legacy `/query` aliases points at `/api/query`, silently moving clients off the path they requested
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 50/100
- Issue 类型
- 缺陷
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- javascript
调研方向
Start with routes/compatability.js, controllers/utils.js, and routes/tests/query.test.js, then inspect how req.originalUrl differs from req.baseUrl + req.path for the three aliases. Confirm the chosen link behavior with a POST test, update public/API.html, and ensure the decision is explicit and covered by the test.
由索引模型根据 Issue 内容生成。
描述
Summary
/v1/api/getByProperties, /v1/api/query.action, and /v1/api/getByProperties.action are URL rewrites onto /v1/api/query (routes/compatability.js:8-10). Now that #312 gives /query a rel="next" Link, a response to any of those aliases carries a next link naming /v1/api/query rather than the path the client actually requested. A client that walks its pages leaves the legacy path after page 1 and is never told.
Verified on production 2026-09-24, read-only:
POST /v1/api/getByProperties?limit=1 {"type":"Annotation"}
POST /v1/api/query.action?limit=1 {"type":"Annotation"}
POST /v1/api/getByProperties.action?limit=1 {"type":"Annotation"}
-> 200, all three:
Link: <https://store.rerum.io/v1/api/query?limit=1&skip=1>; rel="next"
Why this matters
setNextPageLink() builds the path from req.baseUrl + req.path (controllers/utils.js:155). express-urlrewrite replaces req.url before the router matches, so by the time the /query controller runs those two describe the rewritten route. req.originalUrl is untouched and still holds what the client sent, so the alias is recoverable either way.
Pointing at the canonical path is a defensible thing for rel="next" to do. The problem is that it is currently a side effect of how the rewrite works rather than a decision we made, and it happens silently:
- Nothing in the response says the path changed. There is no
Content-Location, so a client cannot know page 2 lives elsewhere without diffing URLs it did not construct. - The aliases exist because old clients still call them. Those are the clients least likely to survive being handed a path their stack was never set up for — fixed base paths, proxy rules, allowlists.
- Page 1 and page 2 of one walk arrive from two different endpoints, so anything that caches, logs, or rate limits per path sees one walk as two.
Scope is narrow: only paged endpoints are affected, which is /query through its three aliases. The /batch_create -> /bulkCreate rewrite does not page.
Neither the behavior nor the aliases are covered anywhere. routes/__tests__/query.test.js never exercises a rewritten path — __tests__/core_provider_contract.test.js:14 only asserts that compatability.js is mounted — and public/API.html does not mention getByProperties at all.
Proposed change
Decide which of these we mean, then make it explicit in code and in public/API.html.
-
Keep the canonical link and announce it. Serve
Content-Location: /v1/api/queryon responses that arrived through a rewrite, so the move from alias to canonical path is stated rather than inferred. One header, and the current behavior stops being accidental. -
Keep the client on the path it asked for. Build the link from the pathname of
req.originalUrlinstead ofreq.baseUrl + req.path. For every non-rewritten route the two are identical, so nothing else changes. This keeps a legacy walk entirely on the legacy path, at the cost of advertising a deprecated URL in a header we control.
Recommend option 1. The canonical path is the better destination, and the only real complaint is that we currently arrive there without saying so.
Either way, add a case to routes/__tests__/query.test.js that POSTs to /v1/api/getByProperties and asserts the rel="next" path, so the choice cannot drift, and give the aliases a line in public/API.html noting that they are rewrites onto /query and what their next links point at.
- 主要语言
- JavaScript
- 星标
- 3
- 派生
- 6
- 平均合并
- 14 分钟
- 30 天内合并 PR
- 2
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
CenterForDigitalHumanities/rerum_server_nodejs 的其他 Issue
-
bug documentation
难度 2/5 1-3 小时 新手友好度 88/100
-
backend dependencies easy
难度 2/5 1-3 小时 新手友好度 84/100
-
难度 2/5 1-3 小时 新手友好度 82/100
CenterForDigitalHumanities/rerum_server_nodejs#290 · 1 条评论 ·
-
难度 3/5 1-2 天 新手友好度 74/100
-
难度 4/5 3-5 天 新手友好度 35/100
查看 CenterForDigitalHumanities/rerum_server_nodejs 的全部 Issue
相似的 Issue
-
check:failed logos:add
难度 2/5 1-3 小时 新手友好度 67/100
iptv-org/database#38874 · 1 条评论 ·
维护者通常 4 天内回复
-
documentation
难度 2/5 1-3 小时 新手友好度 85/100
githubnext/gh-aw-workshop#5193 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 72/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 72/100
-
agent/quality stage/040-covered-by-pr testing
难度 2/5 1-3 小时 新手友好度 85/100
维护者通常 1 天内回复