skip to content

Results & Forwarding Breakage

SPF yields pass, fail, softfail, neutral, none, temperror or permerror, and the receiver picks the action. Interviewers ask -all versus ~all, and why forwarding fails SPF unless SRS rewrites it.

on this pageshow

explore

questions

6

What are the seven results an SPF check can return, and how does a receiving mail server typically handle each one?

level: juniorimportance: must knowfreq 60%

answer

  1. the sender states, the receiver decides
  2. three verdicts, two non-answers, two errors
  3. neutral handled exactly like none
  4. softfail alone is no reason to reject
  5. 451 for transient, 550 for permanent

basics

~20 s

SPF returns pass, fail, softfail, neutral, none, temperror or permerror. RFC 7208 leaves the action to the receiver: it may reject fail, should not reject on softfail alone, must treat neutral like none, and may defer temperror.

solid answer

~50 s

An SPF check compares the connecting IP address with the policy published by the `MAIL FROM` (or `HELO`) domain and returns one of seven results. `pass` means the IP is authorized; `fail` is an explicit statement that it is not; `softfail` is a weaker "probably not". `neutral` means the domain chose to assert nothing, and `none` means there was no usable domain or no SPF record. `temperror` is a transient, usually DNS, error; `permerror` means the record could not be interpreted. RFC 7208 sets almost no handling rules: a receiver may reject `fail` with `550`, SHOULD NOT reject on `softfail` alone, MUST treat `neutral` exactly like `none`, may defer `temperror` with `451`, and may reject `permerror` with `550`. When it accepts a message, it should record the result in a header field for later filters.

go deeper

for a junior

Recall all seven names and one line each: pass and fail are firm statements, softfail is a weak one, neutral and none say nothing, and the two errors are transient or permanent.

for a middle

Explain that RFC 7208 leaves handling to receiver policy, then give the guidance per result: neutral equals none, softfail is not grounds alone for rejection, fail may be rejected with 550.

for a senior

Show operating judgment: what your receiver does with fail and softfail, why accepted mail gets a recorded result, and how forwarding produces fail for legitimate mail.

for a principal

Frame SPF results as one signal in a layered decision: what rejecting at the SMTP stage costs in false positives against what delivering unauthorized mail costs your users.

## What an SPF check produces **Sender Policy Framework** (SPF, RFC 7208, April 2014) lets a domain publish, in a `v=spf1` TXT record, which hosts may send mail using that domain in the SMTP envelope. A receiving mail server runs the evaluation the RFC calls `check_host()` with three inputs: the **IP address of the connecting client**, the **domain** being checked (from `MAIL FROM`, or from `HELO`/`EHLO`), and the sender address. The output is always exactly one of seven named results. The important design choice is in section 2.6: the protocol "establishes no normative requirements for handling any particular result." The domain owner makes a statement; the **receiver's local policy** decides what to do with it. Section 8 then offers guidance per result, and it contains only one MUST: `neutral` MUST be treated exactly like `none`. ## The seven results | Result | What it means | Handling RFC 7208 section 8 suggests | |---|---|---| | `none` | No usable domain could be taken from the session, or the domain publishes no SPF record (a name that does not exist also gives `none`) | No information at all; carry on as if SPF were absent | | `neutral` | A record exists but the domain explicitly declines to say whether this IP is authorized | MUST be treated exactly like `none` | | `pass` | The client is authorized to use the domain in this identity | The domain can, in the sense of reputation, be held responsible for the message; further checks proceed with confidence in that identity | | `fail` | An explicit statement that the client is **not** authorized | Local policy: reject in the transaction with `550` and `5.7.1` (RFC 7372 later registered `5.7.23` for this), or accept and record the result | | `softfail` | A weak statement that the host is **probably not** authorized | SHOULD NOT reject based solely on this result; MAY apply closer scrutiny | | `temperror` | A transient, generally DNS, error during the check; a retry may succeed | Accept, or temporarily reject with `451` and `4.4.3` | | `permerror` | The domain's records could not be interpreted; a DNS operator must fix them | Reject with `550` and `5.5.2`, or accept with an annotation | ## Two classes of handling Whatever the result, a receiver has essentially two kinds of response (section 8): 1. **Act inside the SMTP session** that is delivering the message: a permanent rejection (a `5xx` reply) or a temporary one (a `4xx` reply that asks the sender to try again later). 2. **Accept the message** with a success reply and add a header field, `Received-SPF` or `Authentication-Results`, that records the result for downstream processors such as content filters, a DMARC evaluator or the mail client. RFC 7208 says that a receiver which does not reject SHOULD add that header field, and that this matters most for `fail`, because the message is explicitly not authorized by the domain. Section 11.7 adds a warning: an operator who delivers `fail` mail is admitting content the purported sender has disowned. Receivers genuinely differ. Some treat `fail`, and sometimes `softfail`, as definitive and reject. Others only record the result. Many use SPF as one input among several. Section 10.2 explains why caution is reasonable: SPF can return `pass` for harmful mail (a spammer can publish a correct record for a throwaway domain) and `fail` for legitimate mail that passed through a forwarding alias. ## Reading the results in pairs Most interview confusion disappears when the results are compared two at a time: - **`pass` against `fail`** are both explicit statements by the domain owner. Neither says anything about the message content or about the `From` header the reader sees. - **`fail` against `softfail`** differ in the strength of the owner's statement. `softfail` exists so a domain can say "probably not" without asking receivers to reject; RFC 7208 suggests it might lead to a flag in the mail client or greylisting rather than a refusal. - **`neutral` against `none`** differ only in whether a record exists. The distinction is informational, and section 8.2 explains the MUST: treating `neutral` more harshly would discourage domains from testing SPF. - **`temperror` against `permerror`** differ in who must act. A `temperror` may clear on its own when DNS recovers; a `permerror` persists until the domain's DNS operator corrects the record. ## What a pass does and does not prove - It proves the connecting IP is authorized for the **envelope domain** (or HELO name) that was checked, nothing more. - It does not authenticate the visible `From` header; connecting that header to SPF is what DMARC alignment adds, which is covered with DMARC. - It does not make the message safe: reputation and content checks still apply. - A result is tied to one hop, so mail that passes at the first receiver can fail at a later one after forwarding. ## In practice A mail administrator usually meets these results in logs and in the recorded result field of a delivered message. The working habit is to ask two questions of any result: what did the domain owner state, and what did this receiver's policy choose to do with that statement. Keeping those two apart is exactly what RFC 7208 does.

  • Why does RFC 7208 insist that neutral be treated exactly like none?
    Section 8.2 makes it a MUST so that domains can publish a record that asserts nothing while they work out their sending infrastructure. If receivers punished `neutral` more than having no record at all, domains would be discouraged from deploying and testing SPF. The difference is kept only for information: `neutral` tells you a record exists.
  • If a receiver accepts a message whose SPF check returned fail, what should it do?
    RFC 7208 section 8.4 says it SHOULD add a `Received-SPF` or `Authentication-Results` header field so downstream processors, such as spam filtering, a DMARC evaluator or the mail client, can act on the result. Section 11.7 warns that delivering `fail` mail admits content the purported domain explicitly did not authorize, so it deserves extra scrutiny.
  • Which SMTP replies does RFC 7208 suggest when a receiver does reject on an SPF result?
    For `fail`, `550` with enhanced code `5.7.1`; RFC 7372 (September 2014) registered `5.7.23` "SPF validation failed" for use in its place. For `temperror`, `451` with `4.4.3`, and for `permerror`, `550` with `5.5.2`; RFC 7372's `X.7.24` "SPF validation error" can replace those two.

  • Softfail means the receiver must reject the message.
  • Neutral should be scored more harshly than having no SPF record.
  • An SPF pass proves the visible From address is genuine.
  • RFC 7208 requires receivers to reject every SPF fail.
  • Temperror and permerror both mean the sender's record has a typo.
open as a page

Why does SPF fail for mail forwarded through an alias, such as an alumni address, when the original sender's record is correct?

level: middleimportance: must knowfreq 45%

basics

~20 s

SPF checks the IP of the server that connected last. A plain forwarder resends from its own IP but keeps the original MAIL FROM, so the receiver checks the sender's record, which never listed the forwarder.

open as a page

How does the Sender Rewriting Scheme (SRS) let a mail forwarder pass SPF while bounces still reach the original sender, and why is the rewritten address signed?

level: seniorimportance: should knowfreq 24%

basics

~20 s

SRS rewrites MAIL FROM to an address in the forwarder's own domain that encodes the original sender, so receivers check the forwarder's SPF record. Bounces return there, are verified by hash and timestamp, and are relayed back.

open as a page

An SPF check returns temperror for one sender and permerror for another; what does each mean, and how should the receiving mail server respond?

level: seniorimportance: should knowfreq 20%

basics

~20 s

Temperror is a transient, usually DNS, failure a retry may clear: accept, or defer with 451. Permerror is a record that cannot be interpreted until its DNS operator fixes it: reject with 550, or accept and annotate.

open as a page

For an organisation's SPF records, how do you as the lead decide whether each domain should end in -all or ~all?

level: principalimportance: should knowfreq 30%

basics

~20 s

-all makes unlisted hosts fail; ~all makes them softfail, which RFC 7208 says should not alone cause rejection. Decide per domain from how complete the sender inventory is, how much mail is forwarded, and whether DMARC enforcement and DKIM carry protection.

open as a page

For a bounce with an empty MAIL FROM, which identity does an SPF verifier check, and how does the separate HELO check relate to it?

level: seniorimportance: nice to knowfreq 12%

basics

~20 s

With a null reverse-path, RFC 7208 defines the MAIL FROM identity as postmaster@ the HELO name. The separate HELO check, RECOMMENDED and run first when both are done, evaluates that same host name, so its SPF record decides both.

open as a page