Summary

  • RFC 1036 required an Approved: line for moderated-newsgroup articles and said the moderator should add a mail address, making approval visible to news software.
  • The header was a routing and admission signal, not a cryptographic credential: a later historical draft explicitly said the claimed approval could not be authenticated.

A decision had to travel with the article

Usenet was not one server receiving a final, centrally approved copy. Articles moved between sites that exchanged selected newsgroups and maintained their own local stores. A moderation rule therefore had to survive a journey across software that might not know the moderator personally. The article needed a machine-readable indication of whether it was ready for ordinary propagation or still belonged in a review queue.

RFC 1036, published in December 1987 as the standard for exchanging Usenet messages, gave that indication a short name: Approved. The field appeared in the general list of optional headers, but the rule for moderated groups made it compulsory. The moderator “should” add it, and its value should be the moderator’s mail address. An address was useful: it identified a route back to a responsible person and could be recognized by software. The document did not describe a signature, a key, or a trusted directory against which a receiving site could validate that address.

The wording held two jobs apart, even if readers could easily conflate them. A moderated article was expected to have crossed a human decision point. The header told the software that the decision had been represented in the article. It did not show the review itself, say what standard the moderator applied, or prove who had made the decision. “Approved” named a state; the address named a contact. Neither was a receipt of an authenticated act.

One field covered both posts and group-control messages

RFC 1036’s example gives the marker a concrete shape. A readership summary dated 1 October 1986 appears cross-posted to news.lists and news.groups, with an Approved field naming the moderator’s address. The accompanying Xref records local article numbers on the seismo host. The example shows an address-shaped approval marker traveling as part of a normal news article; it does not tell us how often moderators used the field or how each site checked it.

The header also appeared in a more consequential setting. RFC 1036 describes a newgroup control message that could request creation of a group, including the option to create it as moderated. Sites were told to ignore such a message unless its header included Approved. An rmgroup message, which requested that a group be removed from every host, faced the same requirement. The RFC emphasized care because a removal had network-wide scope. Here the marker helped news software distinguish an ordinary control message from one carrying an approval claim. It still did not identify a cryptographically verified authority.

That reuse exposed a design tension. A message format could make a decision legible across a distributed system without making the decision self-authenticating. Each site could apply the rule that a field must be present, but the field’s presence alone could not establish that the named mailbox belonged to the recognized moderator or group administrator. The protocol supplied a common checkpoint; local arrangements supplied the trust behind it.

The later draft named the trust gap

The history of “Son of 1036” needs careful dating. RFC 1849 was published as a Historic document in March 2010 to preserve a draft that had circulated widely in the early 1990s and, according to its author, had become a de facto reference for news-software implementers. It explicitly says it is historical background rather than current implementation advice.

That retrospective document made the software boundary clearer. A posting agent handled a local person’s submission: it could prepare an article, check whether it was compliant, and pass it toward distribution. A relayer received already prepared articles from other relayers and moved copies onward. For a moderated group, an article without Approved was a reader’s contribution that the posting agent should send to the moderator. A relayer should discard an unapproved article for a moderated group rather than turn it into an email request. The same program might perform both roles, but the responsibilities were not interchangeable.

The draft then stated the limitation directly: there was no way to authenticate an Approved header to ensure the claimed approval had actually been given, and no maintained mechanism to establish a current list of legitimate approvers. It proposed cryptographic authentication as a worthwhile extension, while noting that the problem was difficult on systems without real-time access to key servers. The observation is specific to the draft’s account; it is not evidence that a particular forgery occurred or that every site accepted a forged field.

A later protocol, RFC 5537, formalized the injection path: an injecting agent forwards an unapproved article to the moderator of the leftmost moderated group, customarily by email. Approval and injection remain separate from routine relaying. That later specification shows how the division of labor could be made explicit; it should not be projected unchanged onto every 1980s news installation.

What the marker accomplished—and what it left open

The Approved: line was not meaningless because it lacked cryptographic proof. It gave widely separated news software a shared condition for handling an article and a practical address for moderation. It made an editorial decision portable enough to travel with a distributed publication system. But that portability stopped at the edge of the trust arrangement that surrounded the header.

That distinction is the historical point. The network could carry a claim about approval; it could not infer the authority behind the claim from the text alone. Human review, address administration, local policy and software handling remained different parts of the process. The standard created a place for the decision to appear, not a universal mechanism for proving who had made it.

Sources