Email Deliverability

Understanding the “Precedence:” Header

minute read

Though never formally standardized, the Precedence: header has been around since the earliest days of internet email.

In 1997, RFC 2076 described it as “non-standard, controversial” and “discouraged” — but also gave a clear description of what it was for: “Sometimes used as a priority value which can influence transmission speed and delivery.” At that time, common values were “bulk” and “first-class.” Even in the mid-nineties most email delivery was effectively instantaneous, so the “first-class” designation and influence over speed and delivery faded away (if indeed those ever applied at all.) But “bulk” stuck around.

RFC 3834, titled “Recommendations for Automatic Responses to Electronic Mail”, discusses out-of-office and vacation notices, challenge/response spam filters, autoreply systems intended to distribute information, and so forth. It suggests that automatic responses should not be sent in response to messages with a precedence of “list”, “junk”, or “bulk”, and that the automatic responses should themselves include one of those values in their own Precedence: header to avoid autoresponder loops. While still not providing a formal standard definition, this has long been a best practice and is the most common use of the Precedence: header on the internet today. (It should be noted that “list” usually refers to discussion lists, while “bulk” is more common for one-to-many mailings.)

Commenting on an article on Return Path’s other blog, In the Know, Ram pointed out that if he includes a “Precedence: bulk” header in messages to Gmail users, then Google’s Smart Labels feature will label the message as bulk.

Google’s use of the Precedence: header for labeling is a new idea, and (probably) unique. Since only Google is doing it it can’t be considered a best common practice — it’s not common. But I think it’s a pretty good idea. In the same spirit as other recent uses of the header, this bulk label is simply a loose indicator of the type of message. It doesn’t mean the message is spam; Gmail users can and will decide for themselves how important it is.

Long-time internet email users are already used to setting up automatic user-level filters which shuffle discussion lists and other bulk messages into other folders; I’m quite certain that the lack of these easy, effective filters is one of the leading causes of the “email overload” people keep talking about. Gmail’s bulk label and other Smart Label features give their users just a hint of what’s possible.

(Oh, and if you send bulk mail but don’t label it appropriately with “Precedence: bulk”, I’m certain Gmail’s filters will notice. They may even consider it a sign of spam. Come on, this is one of the few chances you’ll get to tell your recipients and their mail software what your message is all about! Don’t try to lie — lying is what spammers do.)

We should also mention the “bulk” folder in Yahoo! Mail. Yahoo! has for many years had a different filtering policy than the other big webmail systems: rather than blocking messages outright, they preferred to put all suspected spam into the bulk folder. Early on, this was determined by a similarity algorithm which recognized when a substantially similar message was sent to a whole lot of different recipients. Obviously it’s gotten a lot more complicated since then; spammers have been adding randomness for years.

This led to the term “bulking”, which refers to Yahoo!’s bulk folder but also encompasses the “junk” folder at Hotmail, and “spam” folders almost everywhere else. We’ve been scratching our heads, trying to come up with a more descriptive (and less confusing) phrase — any ideas?

RFC 2076 also lists the Priority: and Importance: headers. Both are artifacts of an earlier standardized mail system, X.400, which is often spoken of in hushed tones during IETF meetings. Apparently, X.400 was both the most amazing and the most awful protocol of its’ time; the simplicity of SMTP and internet mail was a direct reaction to X.400’s complexity. Both of those headers (and many others) still appear from time to time, but not often — and they don’t appear to have any real effect today.