[Completed] I add support for Chinese or Unicode names for repositories/users/organizations/teams to Gitea

[znsoooo]

What problem would this solve?

（标题：为Gitea添加支持中文仓库名/用户名/组织名/团队名的功能）

This proposal was rejected by the official (#13943). And the original issue has been closed, I cannot reply there, So I am opening a new issue.

（这是一个已经被官方拒绝的提案（#13943），但是由于原issue已被关闭，我无法回复，所以我新打开一个issue）

Since the proposal has been rejected, I will not submit a PR.

（既然提案已经被拒绝，我不会提交PR）

What do you propose?

I forked the repository code and removed all the code that checks for valid names. Now it supports Unicode for usernames, repository names, organization names, team names, and badge names.

（我fork了仓库代码，通过移除了所有关于命名合法性校验的代码，现在实现支持中文和Unicode用户名、仓库名、组织名、团队名、和徽章名）

Using effect (使用效果):

.

Detail introduction and Windows executable download link (具体介绍和Windows可执行程序下载):

https://github.com/znsoooo/gitea-support-chinese-repo-name

Related issues (相关issue):

• #13943: Allow "+" (plus sign) in a repository name
• #18405: Additional Non-English names for repository, organization and teams
• #18816: Allows create organization name, Repositories name, and teams name in Local character set
• #13295: Support different/localized language/character/letter for organization/repository name
• #19699: Can the repository name support UTF-8 names? i.e. Chinese
• #18832: Add nickname for repo
• #9032: Add human readable project name at repo settings page

People who might be interested (可能感兴趣的人):

@andrewlukoshko @bigsk05 @clarfonthey @D0ctorWh0 @davidlt @dxf1122 @kevinlin18 @myoung1032 @omeryildiz58 @rcdevice @sergey-dryabzhinsky @techknowlogick @usernamedd

@delvh @lunny @silverwind @wxiaoguang @zeripath

@newhinton @jolheiser @lafriks

[lafriks]

I don't really see reason to support this as that will just create a lot of edge cases and problems with path handling and underlying filesystems

[znsoooo]

@lafriks I have read through all your discussion. I respect your decision not to add this feature, but I also hope you understand that some people need it, including me.

Possible scenarios where it might be needed:

• Some projects are named without considering an English name; they only have a personalized Chinese name. In such cases, translating the project name into English becomes a disaster, whether by free translation or pinyin.
• The registered users are also Chinese. Using pinyin would be hard to understand, and using an English name would make it even less recognizable.
• Sometimes we manage the toolchain of partner companies, and those companies also have Chinese names. If Chinese is not supported, I can only use an unfamiliar English company name or pinyin, neither of which is ideal.
• Our organizational unit is called "一部二组". How should I translate that into English? "Department-1-Group-2"? Isn't that too long?
• People visiting the project page on the web are not developers; they just need to read the README document, or download several files. They only know the Chinese name of the project. If all paths need to be translated into English/pinyin, it becomes difficult for them to find things.

And your biggest concern for rejecting this proposal is cross-platform compatibility:

But we don't operate across platforms at all. We found one computer, set up a code hosting platform, and provide it for employees to use. Why would we need to migrate from Windows to Linux, and back, and again, and again?

[lafriks]

Problem is that it would be large change as it's not just single validation regex to change, it would take a lot of other changes to support this, including a lot of CI/CD and other tool integrations that would most probably break etc. We can not just add exception for one limited usecase as there are a lot of gitea instances that are not used in that limited scope

[newhinton]

Honestly, i do understand @znsoooo arguments. This is an accessibility issue.

But maybe there is a middle ground:

The problem isn't actually the urls or paths or filesystem used by gitea, but the fact that it is the same than the displayname. Keep the plumbing, but decouple the displayname/groupname from the actual used path and the problem should be solved.

I don't know how hard that is to pull off, but i imagine quite a bit easier than rebuilding the underlying infrastructure.

[znsoooo]

Problem is that it would be large change as it's not just single validation regex

@lafriks I said I respect your team's choice, so I submitted a fork, for someone who needs it to download.
(I saw you clicked a dislike, seems like you have strong opinions about this issue)

---

including a lot of CI/CD and other tool integrations that would most probably break

I haven't done related testing. Let me share a fact that might surprise you: our company doesn't have CI/CD (we're not a programming-focused company). Gitea's ease of deployment has attracted many non-programmers company, so supporting Unicode repository names could further​ improve its usability.

At the very least, this fork should be allowed to exist.

Alternatively, there's a middle ground: this could be made a config option, validation enabled by default, but can be turned off to skip those checks.

[wxiaoguang]

Just a kind reminder, your blind change znsoooo@c0b45f4 will definitely cause problems.

For example, you need to take care of:

• names from attackers like "../../"
• names which will conflict "name.git/"

Anyway, that's your choice, take your own risks. (btw: no need to mention me in the discussion, I won't discuss more for your changes)

[znsoooo]

For example, you need to take care of:

• names from attackers like "../../"
• names which will conflict "name.git/"

@wxiaoguang This is a simplified way of saying "I removed ALL the checks for validation", but actually I still kept some checks, like checking for ".git" or ".wiki" suffixes, like blocking reserved words for ".." or ".".

The known problem is including "/" or "\" in the repository name. (It will return a 500 error, or the creation succeeds but access fails)

But our server is set up on our company's computers, provide services for company's employees. If an attacker in our company, we'll deal with that attacker.

---

Update:

If "../../" exists in a repository name, it could cause the repository folder to be created outside the gitea program directory, that is not a good result. Perhaps I should relax the validation, but still keep the rule that repository names and usernames cannot contain "/" or "\". This might be easy to achieve by modify the pattern.

However, considering that even after making these changes, your team still would not accept the feature, so I'm not planing continue to modify it.

BUT if someone who AGREE with the feature that using Chinese repository names, and they wants this feature, I will consider build another one.

[lafriks]

As I said that if in your usecase that works it is fine and just patch upstream gitea and deploy your own patched version but such changes are just too breaking and will be hard to support here. That is the beauty of open source

[silverwind]

We can expose regexes in config for repo/user/org/team names, but the defaults should stay compatible with GitHub/GitLab to allow easy migrations. That said, I think there are still some divergences that should be fixed.

[lafriks]

They can be exposed but only as unsupported feature that if there are bugs and/or something is not working you are on your own and will not be fixed in Gitea code, otherwise it will be neverending corner cases and mess in the code to support them

[silverwind]

Yes, these options should have a warning disclaimer attached.

[znsoooo]

We can expose regexes in config for repo/user/org/team names, but the defaults should stay compatible with GitHub/GitLab

I agree.

In addition, I remember that the team names / badge names only exist in the database, right? Why they also to be set "AlphaDashDot" only?