This site's argument is that the software publisher can selectively attack users during a live software install, in a way that they don't stand a chance of detecting by inspection (or of having proof of after the fact).
1) Distributing software via bash script is a bad idea
2) Sensible people review the bash scripts they downloaded before running them
3) But haha! Here is a clever trick that evades that review.
And I'm not persuaded by 3) being interesting because I already rejected 1) and 2), and I consider 3) to just be proving my point -- you (for all you!) are not competent to perform a very brief but somehow thorough security review of a shell script that probably has further dependencies you aren't even looking at, and the actual reasoning to apply when deciding to install software this or any way is purely "Do I trust the entity I have this TLS connection open with to run code on my machine?".
I think there's a difference between trusting an organization's code that is published to the general public, and trusting an organization to send you arbitrary code in a specific moment. Only software distribution methods can enforce this kind of distinction, and curl | bash by itself doesn't, particularly in light of the article's technique.
I tried to discuss this distinction in some of my reproducible builds talks. There's a difference between trusting Debian to publish safe OS packages, and trusting Debian to send you a safe package when you run a package manager if the package could easily be different every time. This is particularly so when someone may be able to compromise the software publisher's infrastructure, or when a government may be able to compel the software publisher to attack one user but other users.
Instead of your (1) and (2) above, how about this?
1) Distributing software via a method that can single out particular users and groups to receive distinctive versions is a bad idea: it increases the chance that some users will actually be attacked via the software publication system.
2) We might think that curl | bash isn't particularly egregious this way, because there are various ways that publishers might get caught selectively providing malicious versions. This is especially so because the publishers can't tell whether a curl connection is going to run the installer or simply save it to disk. That makes the publishers (or other people who could cause this attack to happen) less likely to do it.
3) But haha! Here is a clever trick that restores the publishers' ability to distinguish between running and saving the installer, and in turn breaks the most plausible ways that publishers could get caught doing this.
Edit: Elsewhere in this thread you suggested that the likeliest alternative is something like
curl https://somesite.com/apt.key | sudo apt-key add - && sudo apt-get update && sudo apt-get install some-software
I think I'd agree that this has some of the same problems, although it might have some advantages because of the potential decoupling between the distribution of the signing key and the distribution of the signed package. As another commenter pointed out, you could try to use a different channel to get or verify the key, and some users actually do; also, you'll have a saved copy of the key afterward.
I was assuming that the sites that you might `curl | bash` from are third-party sites (i.e. not your Linux distribution) that you don't have an existing trust relationship with, which makes it impossible to avoid this capability. That's the situation people use curl | bash in.
So I think this ability to individualize artifacts would still be present if we were receiving a .deb or apt key instead from that site.
> you'll have a saved copy of the key afterward
Yes, though since dpkg post-install scripts can modify arbitrary files (right?), you can't trust that any files on your disk are the ones that existed before the compromise. So couldn't the malicious key verify the malicious package, which then overwrites the copy of the package and key on-disk with the good versions that were given to everyone else?
I guess we need some other infrastructure or social practice on top in order to compare what different people see, and/or allow the distributor to commit to particular versions. (Then having the distributor not know whether someone is blindly installing a particular file without verification is necessary, but not sufficient, to deter this kind of attack.)