Ask HN:为小型团队托管 Fossil – 有趣,还是错误的选择?

4作者: ragelink7 天前
我几个月来一直在开发一个托管的 Fossil SCM 服务,我真的不知道这是否是个好主意。今天首页上的“我们需要一个锻造厂联盟”的帖子让我觉得值得发出来分享一下。<p>我正在构建的内容:一个托管 Fossil 仓库的家。与代码托管平台有相同的入门体验,但每个项目都是一个可以克隆、通过电子邮件发送或带走的独立的自包含 SQLite 文件。开源的综合平台(Django + Postgres + Redis + Caddy + Litestream-to-S3)位于 fossilrepo.io。托管版本很快就会进入私人测试阶段。<p>我的粗略论点: 1. Fossil 已经通过设计实现了联邦化。每个克隆都是整个项目:问题、维基、论坛、历史记录、代码。这正是首页上正在讨论的联邦化问题,只是使用了一个已有 15 年历史的工具,SQLite 项目本身也在使用。如果 fossil clone 在任意两个主机之间都能工作,锁定效应基本上就消失了。<p>2. AI 代理需要集成的上下文。一个 Fossil 仓库就是一个可查询的 SQLite 文件。一个代理使用 SELECT * 读取代码 + 票据 + 维基 + 历史记录,而不是 47 个左右的 GraphQL 调用。RAG 和 MCP 设置变得微不足道。它还有一个超级容易使用的 cli 工具。<p>3. 小型但认真的团队没有得到充分的服务。Git+GitHub 赢得了宏观市场,而且这种情况不会改变。但是对于 1-50 人的团队来说,规范、票据、维基和代码都应该放在同一个地方……集成模型更好/更容易。<p>我担心的事情: - 网络效应——如果其他人找不到仓库,它就没什么用 - 惯性——Git 的肌肉记忆很难打破 - Codeberg / Forgejo / Gitea 都是可靠的 — 哪个才是正确的切入点,如果有的话?这是不是一个有人想要解决的方案?<p>我宁愿现在就在 HN 上听到大家的意见,而不是在发布之后。三个诚实的问题: - 这有意思吗,还是我解决了一个没人遇到的问题? - 什么会让你真正切换? - 我遗漏了什么?<p>另外,还在纠结名字。这两个域名都指向同一个页面,所以请投票选择你实际会使用的那个:https://fossilforge.io 或 https://fossilhub.io(我也有 .ai 版本,但 .io 感觉更像开发者)
查看原文
I&#x27;ve been working on a hosted Fossil SCM service for a few months and I genuinely don&#x27;t know if it&#x27;s a good idea. The &quot;We need a federation of forges&quot; thread on the front page today made me think it&#x27;s worth posting.<p>What I&#x27;m building: a hosted home for Fossil repos. Same onramp feel as a code host, but each project is a single self-contained SQLite file you can clone, email, or walk away with. The open source omnibus (Django + Postgres + Redis + Caddy + Litestream-to-S3) is at fossilrepo.io. The hosted version will be in private beta soon.<p>My rough thesis: 1. Fossil is already federated by design. Every clone is the entire project: issues, wiki, forum, history, code. That&#x27;s the federation discussion happening on the front page right now, just with a 15+ year-old tool the SQLite project itself uses. If fossil clone works between any two hosts, lock-in basically dies.<p>2. AI agents need integrated context. A Fossil repo is one queryable SQLite file. An agent reads code + tickets + wiki + history with SELECT * instead of 47-odd GraphQL calls. RAG and MCP setups become trivial. Also has a cli tool thats super easy to use.<p>3. Small-but-serious teams are underserved. Git+GitHub won the macro market and that&#x27;s not changing. But the 1-50 person team where the spec, the tickets, the wiki, and the code all belong in the same place... the integrated model is just better&#x2F;easier.<p>Things I&#x27;m worried about: - Network effect — a repo isn&#x27;t very useful if nobody else can find it - Inertia — Git muscle memory is hard to break - Codeberg &#x2F; Forgejo &#x2F; Gitea are all credible — what&#x27;s the right wedge, if any? is this a solution that anyone wants?<p>I&#x27;d rather hear it from HN now than after we launch. Three honest questions: - Is this interesting, or am I solving a problem nobody has? - What would make you actually switch? - What am I missing?<p>Also stuck on the name. Both domains land on the same page, so vote whichever you&#x27;d actually use: https:&#x2F;&#x2F;fossilforge.io or https:&#x2F;&#x2F;fossilhub.io (I also have the .ai versions but .io feels more developer-y)