$ go version go version devel +bdf690b6a741 Fri Mar 08 10:41:20 2013 -0800 linux/386 $ go tool | grep '^.g$' 8g $ uname -a Linux dragon 3.5.7-gentoo-devtmpfs #2 SMP Fri Feb 1 13:03:07 EST 2013 i686 Pentium III (Katmai) GenuineIntel GNU/Linux $ I expect the following command to go get a package. What it does instead is shown below. $ go get -v sethwklein.net/go/gogettest Fetching https://sethwklein.net/go/gogettest?go-get=1 https fetch failed. Fetching http://sethwklein.net/go/gogettest?go-get=1 Parsing meta tags from http://sethwklein.net/go/gogettest?go-get=1 (status code 200) import "sethwklein.net/go/gogettest": parse http://sethwklein.net/go/gogettest?go-get=1: no go-import meta tags package sethwklein.net/go/gogettest: unrecognized import path "sethwklein.net/go/gogettest" $ Note it saying, "no go-import meta tags". Let's check... $ curl -i 'http://sethwklein.net/go/gogettest?go-get=1' HTTP/1.1 200 OK Date: Fri, 08 Mar 2013 18:51:03 GMT Server: Apache Last-Modified: Fri, 08 Mar 2013 18:48:39 GMT ETag: "cdfd0-b3-4d76e457c6fc0" Accept-Ranges: bytes Content-Length: 179 Content-Type: text/html $ Looks like a go-import meta tag to me. Now if that was served as xhtml, I could think of four or five problems with it just off the top of my head. But it's not. It's served as text/html with no file extension. This version works: $ go get -v sethwklein.net/go/gogettestworks Fetching https://sethwklein.net/go/gogettestworks?go-get=1 https fetch failed. Fetching http://sethwklein.net/go/gogettestworks?go-get=1 Parsing meta tags from http://sethwklein.net/go/gogettestworks?go-get=1 (status code 200) get "sethwklein.net/go/gogettestworks": found meta tag main.metaImport{Prefix:"sethwklein.net/go/gogettestworks", VCS:"git", RepoRoot:"http://github.com/sethwklein/gogettest"} at http://sethwklein.net/go/gogettestworks?go-get=1 sethwklein.net/go/gogettestworks (download) sethwklein.net/go/gogettestworks $ curl -i 'http://sethwklein.net/go/gogettestworks?go-get=1' HTTP/1.1 200 OK Date: Fri, 08 Mar 2013 18:53:00 GMT Server: Apache Last-Modified: Fri, 08 Mar 2013 18:48:42 GMT ETag: "cdfcd-ba-4d76e45aa3680" Accept-Ranges: bytes Content-Length: 186 Content-Type: text/html $ The difference is href=foo.css -> href="foo.css". Both versions are valid according to http://www.w3.org/TR/html-markup/syntax.html#attr-value-unquoted and gorgeous compared to much of the html in the wild. I would blame this on using an xml parser to parse html[1], but the presence meta charset tag, also with an unquoted attribute value, in both versions suggests it's not so simple. Thanks to kisielk on #go-nuts for figuring out the problem[2]. [1]: http://golang.org/src/cmd/go/discovery.go?h=parseMetaGoImports#L22 [2]: https://botbot.me/freenode/go-nuts/msg/2224418/