In the old days that wasn't a problem - when you released some OSS, you put a tarball somewhere and told people to download it. Some of those people were redistributors (mostly Linux distros or BSDs, but also CD vendors, people running FTP sites of neat things, etc.). Many of those redistributors, in turn, got their software to people from further redistributors (mirrors, people burning CDs and passing them around, etc. - up to even a decade ago I was both giving and receiving Ubuntu CDs). More people got your software from a redistributor than directly from you.
If you tried to "un-release" some OSS, not only would that not work, it would be abundantly clear to you that it wouldn't, and moreover that would be clear to you well before you even published the first version, so you wouldn't feel like you were tricked or didn't have the chance to think this through.
With the new world of GitHub and NPM and such and especially tools like Go that pull directly from GitHub, the role of the distributor is basically n more than the role of GeoCities: they provide hosting for you, but it's in your account. So you can take things out of GitHub and NPM just as easily as you can un-publish a web page. (In fact an analogy could be made here to blogs, where you can un-publish, vs. newsgroups/mailing lists, where you can't.) There could be mirrors, certainly, but there aren't necessarily mirrors, and the social norms on both ends are against them: you aren't explicitly asking people to mirror your code and distribute it independently from you, and other people may feel that it is rude/inappropriate to continue to distribute code that you've chosen to stop distributing.
There are certainly very strong advantages in scale of participation (and in the loss of a certain gatekeeping that could not scale) in this new world, but it does seem like it would be good to recapture this one feature of the old world.
If you tried to "un-release" some OSS, not only would that not work, it would be abundantly clear to you that it wouldn't, and moreover that would be clear to you well before you even published the first version, so you wouldn't feel like you were tricked or didn't have the chance to think this through.
With the new world of GitHub and NPM and such and especially tools like Go that pull directly from GitHub, the role of the distributor is basically n more than the role of GeoCities: they provide hosting for you, but it's in your account. So you can take things out of GitHub and NPM just as easily as you can un-publish a web page. (In fact an analogy could be made here to blogs, where you can un-publish, vs. newsgroups/mailing lists, where you can't.) There could be mirrors, certainly, but there aren't necessarily mirrors, and the social norms on both ends are against them: you aren't explicitly asking people to mirror your code and distribute it independently from you, and other people may feel that it is rude/inappropriate to continue to distribute code that you've chosen to stop distributing.
There are certainly very strong advantages in scale of participation (and in the loss of a certain gatekeeping that could not scale) in this new world, but it does seem like it would be good to recapture this one feature of the old world.