Consider mitigating against typosquatting

[tadasant]

Seeing as we aren't serving the use case of direct downloading of packages (like some npm install <packagename> flow), typosquatting is not a particularly notable risk.

However, we should consider blocking publication of packages within a certain edit distance of existing packages (esp. popular ones).

We may want to use #95 in the definition of "popular".

[joelverhagen]

Typo squatting / homograph spoofing / Unicode normalization is a somewhat deep subject. My guess is that the deeper intent is to mitigate both search (e.g. accent insensitive) and browsing (visual similarity) attacks.

Are there rules/recommendations on how punycode MCP server names (reverse DNS names) are rendered? This is a vector for attack if tools handle the punycode.

[connor4312]

I would say we should restrict to some reasonable subset of printable ASCII characters for server names, and that the registry should use the punycoded version of hostnames to avoid domain confusion.