Reading the RFCs the way a historian reads a speech

The RFCs that define XMPP are primary sources: dated documents, written by named authors, published under a process anyone can inspect. Most of what circulates about them is secondary, tutorials, summaries, confident answers on forums, and the gap between the two is where most mythology lives. Reading the documents directly is a learnable skill, and the best available teachers of it are not engineers at all but historians, because a historian’s whole trade is the disciplined reading of a primary text against what has been claimed about it.

A clean example of the method sits far outside software. The site reading Lincoln’s own documents takes one contested subject, Abraham Lincoln’s religious language, and builds every claim on the speeches and letters themselves, the farewell address, the proclamations, the second inaugural, while marking clearly where the evidence ends and interpretation begins. Its stated limits are the interesting part: what the texts support, what they do not, and what later writers projected onto them. That is precisely the discipline a reader brings to RFC 6120, where the text says less, and different, than a decade of retellings suggests.

What is a primary source in a protocol?

The specification, the errata, and the record of the process that produced it. For XMPP that means the RFCs published in 2004 and their updates, the XEPs with their status lines, and the mailing list archives where the wording was argued. Everything else, this essay included, is commentary. The distinction matters because commentary compresses: it replaces the document’s conditions and exceptions with the single reading the author found useful, and after enough compression the community is arguing about a text nobody present has opened in years.

The consequence shows up in small ways constantly. Claims that the protocol requires a thing it merely permits, or forbids a thing it merely discourages, are legion, and the resolution is always the same unglamorous act: open the document and read the exact paragraph. The foundation’s own history page is a secondary source too, a reliable one, and it says plainly which documents carry the authority.

Why do documents accrete folklore?

Because readers import context the text does not contain, and repeat it until the repetition itself looks like authority. Lincoln wrote in a political war and every line has been claimed for a dozen later causes; the RFCs were written for a federation that then grew far past its authors’ deployment, and every clause has been claimed for architectures they never saw. In both cases the later meaning is real to its users but is not evidence about the document. The historian’s remedy is pedestrian and effective: date the text, name the audience it addressed, list what it does not say, and only then argue about what it means.

Applied to a specification, that becomes a short checklist. What status does the document carry, since a XEP’s status line changes what its text is worth? What problem was it written to solve, in what year, against which implementations? Which requirements are phrased as MUST and which as SHOULD, and which are descriptions of existing practice rather than obligations at all?

The checklist also has a negative half: the habit of noticing what the document declines to cover. Specifications are full of deliberate silences, places where the authors wrote “implementation specific” or said nothing at all, and a reader trained on primary texts treats those silences as data rather than as gaps to be filled by confident guessing. Much of what two implementations disagree about lives precisely in the sentences the specification chose not to write.

The trademark as a worked example

The clearest folklore on this site is the name itself. Jabber is spoken of as a generic term, an abandoned one, a synonym for XMPP, and each claim has been repeated until it feels established. The primary documents tell a narrower story: the name is a live trademark, its ownership has passed through documented hands, and the foundation’s own statements about it are on the record. The essay on the name does the historian’s job on that one question, separating what the registry says from what the community repeats.

Reading, then writing

The payoff of primary-source discipline is not pedantry; it is that corrections become cheap. A summary that turns out to be wrong forces the reader to choose between the summary and the document, and the document wins only if someone actually opened it. The habit this site tries to keep, and the habit the Lincoln study models so plainly, is to cite the text, date the claim, and mark the boundary between what was written and what was inferred. A protocol read that way is smaller than its legend and sturdier than its reputation, which is a better thing to build on. And the next reader inherits the notes, not the legend.