How does 'go get' work when given a not known code hosting domain site?

Let's take an example. The command below will do:

go get robpike.io/ivy

This will get me the content of the repository under $GOPATH/src. Great!

Now, how does it work?

First, robpike.io/ivy replies with a HTTP-redirect:

HTTP/1.1 302 Found

<a href="http://godoc.org/robpike.io/ivy">Found</a>

From reading at go help importpath, I learn that:

> If the import path is not a known code hosting site and also lacks a
> version control qualifier, the go tool attempts to fetch the import
> over https/http and looks for a <meta> tag in the document's HTML

However, looking for a metatag inside the content of the redirected page using:

curl -D --raw https://godoc.org/robpike.io/ivy | grep go-import

returns nothing.

Reading further:

> The repo-root is the root of the version control system containing a
> scheme and not containing a .vcs qualifier.
>
> For example,
>
> import "example.org/pkg/foo"
>
> will result in the following requests:
>
> https://example.org/pkg/foo?go-get=1 (preferred)
>
> http://example.org/pkg/foo?go-get=1 (fallback, only with -insecure)

Again:

curl -D --raw https://robpike.io/ivy?go-get=1

returns nothing.

So the question is: how can I do the same thing as Mr. Rob Pike and use my own site with the go get command?

The command curl -D --raw 'https://robpike.io/ivy?go-get=1' returns an HTML document containing the tag

<meta name="go-import" content="robpike.io/ivy git https://github.com/robpike/ivy.git">

The go get command uses this meta tag to resolve the vanity import path to an actual git repository. You can do the same.

That last command you entered does return data. When I run curl -D --raw https://robpike.io/ivy\?go-get\=1 in my terminal, I get the following data back:

<!-- language: lang-html -->

<meta name="go-import" content="robpike.io/toy git https://github.com/robpike/toy.git"><meta name="go-import" content="robpike.io/cmd/translate git https://github.com/robpike/translate.git"><meta name="go-import" content="robpike.io/cmd/freq git https://github.com/robpike/freq.git"><meta name="go-import" content="robpike.io/cmd/hira git https://github.com/robpike/hira.git"><meta name="go-import" content="robpike.io/cmd/kana git https://github.com/robpike/kana.git"><meta name="go-import" content="robpike.io/cmd/kata git https://github.com/robpike/kata.git"><meta name="go-import" content="robpike.io/nihongo git https://github.com/robpike/nihongo.git"><meta name="go-import" content="robpike.io/cmd/typo git https://github.com/robpike/typo.git"><meta name="go-import" content="robpike.io/filter git https://github.com/robpike/filter.git"><meta name="go-import" content="robpike.io/cmd/unicode git https://github.com/robpike/unicode.git"><meta name="go-import" content="robpike.io/cmd/doc git https://github.com/robpike/doc.git"><meta name="go-import" content="robpike.io/cmd/scrub git https://github.com/robpike/scrub.git"><meta name="go-import" content="robpike.io/cmd/strings git https://github.com/robpike/strings.git"><meta name="go-import" content="robpike.io/ivy git https://github.com/robpike/ivy.git"><meta name="go-import" content="robpike.io/cmd/now git https://github.com/robpike/now.git">

Ths allows the go get command to resolve the vanity path to git repositories.