About two days ago, @freemo made a well-considered proposal to create a collective among Fediverse instances, for the purpose of enforcing respect for one another's freedom of speech. It doesn't quite line up with my idea of free speech, however, and I decided to publish my thoughts on the matter to solicit feedback. Just as "your right to swing your arm leaves off where my right not to have my nose struck begins," so too one person's right to speak leaves off where another's right not to repeat his words begins.

When you access a webpage, your browser sends a request to the server on which that website is hosted. In the simplest case, this request is essentially a formalised way of saying, "Hey, [server], please send me [page]." If things go according to plan, the server responds with the named page. You can set this up yourself: type up some text, share it using a tool like SimpleHTTPServer, and load it in a browser.

But you don't, strictly speaking, *need* to run a server to send the content in response to client requests. You could, for instance, run netcat listening on an appropriate port, and manually type in whatever you wanted as the content of the webpage. Of course, this would get very tedious if you had to type it in more than once or twice, but it does cause the browser to display the content.

It should be evident that typing the response directly back to the listener is an instance of speech, the same way it is to answer a question on the telephone or by a printed letter. In this case, it's very clear who the speaker is - if the reply is based on information obtained from a third party, or even if it's a direct quote of a third party, the role of speaker still remains with the person actually making the reply by typing it up, by speaking it into the handset, or by putting pen to paper.

Note that the freedom of speech allows one person to decline to quote another as well. If Alice asks Bob, "What did Carol say about me?" that freedom allows him to say, "It doesn't bear repeating," or even lie and say, "Nothing at all." It is not a violation of Carol's freedom of speech for Bob to do so, even if Bob knows full well what Carol said about Alice, even if Carol made her remarks directly to Bob, and even if Carol wanted Bob to pass the remarks on to Alice.

Let's return to the idea of manually typing out your HTTP responses. Suppose, for instance, you have a comment section on the page you want to show the user - why would you be obliged to type out a given comment, simply because the commenter wants you to pass it on to your site's visitor? Clearly, as you're the speaker, your freedom of speech is still in operation as in the above scenario, and it protects your decision to decline to do so.

Consider an old-fashioned printing press, which works by pressing an inked design against a sheet of paper. When that paper is retrieved from the machine, it may have some words on it - but the printing press itself is not the speaker of those words. It is the operator of the press - who carved the design, or constructed it from movable type or some other means - who is responsible for their content. The press itself is merely a tool that enables him to produce many copies of the same words more rapidly than he could by handwriting them repeatedly.

Likewise, a webserver is just a tool that saves the site owner from having to repeatedly type the same response to each HTTP request he gets. This is just a program running on a machine somewhere, which itself doesn't have the degree of autonomy necessary to enjoy freedom of speech. The program's speech cannot be "free" in any meaningful sense, because it is completely controlled by the person with the capability to determine how it works, whether by altering the program itself or by modifying the content files it sends.

This is how federation works among Mastodon and related technology. Each instance is normally responsible for passing along the content produced by others, but silencing, suspending, and banning allow the operator of each website to specify policies under which his machine should not repeat that content. This control makes the operator the speaker to whom freedom of speech applies, even though he neither authored the content nor manually transmitted it to the client. His free speech rights, far from compelling him to never use these tools, are in fact what allows him to exercise his discretion to do so.

Although they use tools to distribute content more efficiently, the human operators remain the speakers, and it is to them that the freedom of speech applies. Machines themselves don't have such a freedom protecting them from our decisions to modify the content they emit, and if others offer comments, it is no infringement of their freedom when we decline to repeat their words.