MIME Sniffing

1. Introduction

The HTTP Content-Type header field is intended to indicate the MIME type of an HTTP response. However, many HTTP servers supply a Content-Type header field value that does not match the actual contents of the response. Historically, web browsers have tolerated these servers by examining the content of HTTP responses in addition to the Content-Type header field in order to determine the effective MIME type of the response.

Without a clear specification for how to "sniff" the MIME type, each user agent has been forced to reverse-engineer the algorithms of other user agents in order to maintain interoperability. Inevitably, these efforts have not been entirely successful, resulting in divergent behaviors among user agents. In some cases, these divergent behaviors have had security implications, as a user agent could interpret an HTTP response as a different MIME type than the server intended.

These security issues are most severe when an "honest" server allows potentially malicious users to upload their own files and then serves the contents of those files with a low-privilege MIME type. For example, if a server believes that the client will treat a contributed file as an image (and thus treat it as benign), but a user agent believes the content to be HTML (and thus privileged to execute any scripts contained therein), an attacker might be able to steal the user’s authentication credentials and mount other cross-site scripting attacks. (Malicious servers, of course, can specify an arbitrary MIME type in the Content-Type header field.)

This document describes a content sniffing algorithm that carefully balances the compatibility needs of user agent with the security constraints imposed by existing web content. The algorithm originated from research conducted by Adam Barth, Juan Caballero, and Dawn Song, based on content sniffing algorithms present in popular user agents, an extensive database of existing web content, and metrics collected from implementations deployed to a sizable number of users. [SECCONTSNIFF]

2. Conformance requirements

The keywords "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119. For readability, these keywords will generally not appear in all uppercase letters. [KEYWORDS]

Requirements phrased in the imperative as part of algorithms (such as "strip any leading space characters" or "return false and abort these steps") are to be interpreted with the meaning of the keyword used in introducing the algorithm.

Conformance requirements phrased as algorithms or specific steps can be implemented in any manner, so long as the end result is equivalent. In particular, note that the algorithms defined in this specification are intended to be easy to understand and are not intended to be performant.

3. Terminology

This specification depends on the Infra Standard. [INFRA]

An HTTP token code point is U+0021 (!), U+0023 (#), U+0024 ($), U+0025 (%), U+0026 (&), U+0027 ('), U+002A (*), U+002B (+), U+002D (-), U+002E (.), U+005E (^), U+005F (_), U+0060 (`), U+007C (|), U+007E (~), or an ASCII alphanumeric.

This matches the value space of the token token production. [HTTP-1.1]

An HTTP quoted-string token code point is U+0009 TAB, a code point in the range U+0020 SPACE to U+007E (~), inclusive, or a code point in the range U+0080 through U+00FF (ÿ), inclusive.

This matches the effective value space of the quoted-string token production. By definition it is a superset of the HTTP token code points. [HTTP-1.1]

A binary data byte is a byte in the range 0x00 to 0x08 (NUL to BS), the byte 0x0B (VT), a byte in the range 0x0E to 0x1A (SO to SUB), or a byte in the range 0x1C to 0x1F (FS to US).

A whitespace byte (abbreviated 0xWS) is any one of the following bytes: 0x09 (HT), 0x0A (LF), 0x0C (FF), 0x0D (CR), 0x20 (SP).

A tag-terminating byte (abbreviated 0xTT) is any one of the following bytes: 0x20 (SP), 0x3E (">").

Equations are using the mathematical operators as defined in [ENCODING]. In addition, the bitwise NOT is represented by ~.

4. MIME types

4.1. MIME type representation

A MIME type represents an internet media type as defined by Multipurpose Internet Mail Extensions (MIME) Part Two: Media Types. It can also be referred to as a MIME type record. [MIMETYPE]

Standards are encouraged to consistently use the term MIME type to avoid confusion with the use of media type as described in Media Queries. [MEDIAQUERIES]

A MIME type’s type is a non-empty ASCII string.

A MIME type’s subtype is a non-empty ASCII string.

A MIME type’s parameters is an ordered map whose keys are ASCII strings and values are strings limited to HTTP quoted-string token code points. It is initially empty.

4.2. MIME type miscellaneous

The essence of a MIME type mimeType is mimeType’s type, followed by U+002F (/), followed by mimeType’s subtype.

A MIME type is supported by the user agent if the user agent has the capability to interpret a resource of that MIME type and present it to the user.

Ideally this would be more precise. See w3c/preload #113.

4.3. MIME type writing

A valid MIME type string is a string that matches the media-type token production. In particular, a valid MIME type string may include parameters. [HTTP-SEMANTICS]

A valid MIME type string is supposed to be used for conformance checkers only.

A valid MIME type string with no parameters is a valid MIME type string that does not contain U+003B (;).

4.4. Parsing a MIME type

To parse a MIME type, given a string input, run these steps:

1. Remove any leading and trailing HTTP whitespace from input.

2. Let position be a position variable for input, initially pointing at the start of input.

3. Let type be the result of collecting a sequence of code points that are not U+002F (/) from input, given position.

4. If type is the empty string or does not solely contain HTTP token code points, then return failure.

5. If position is past the end of input, then return failure.

6. Advance position by 1. (This skips past U+002F (/).)

7. Let subtype be the result of collecting a sequence of code points that are not U+003B (;) from input, given position.

8. Remove any trailing HTTP whitespace from subtype.

9. If subtype is the empty string or does not solely contain HTTP token code points, then return failure.

10. Let mimeType be a new MIME type record whose type is type, in ASCII lowercase, and subtype is subtype, in ASCII lowercase.

11. While position is not past the end of input:

    1. Advance position by 1. (This skips past U+003B (;).)

    2. Collect a sequence of code points that are HTTP whitespace from input given position.

       This is roughly equivalent to skip ASCII whitespace, except that HTTP whitespace is used rather than ASCII whitespace.

    3. Let parameterName be the result of collecting a sequence of code points that are not U+003B (;) or U+003D (=) from input, given position.

    4. Set parameterName to parameterName, in ASCII lowercase.

    5. If position is not past the end of input, then:

       1. If the code point at position within input is U+003B (;), then continue.

       2. Advance position by 1. (This skips past U+003D (=).)

    6. If position is past the end of input, then break.

    7. Let parameterValue be null.

    8. If the code point at position within input is U+0022 ("), then:

       1. Set parameterValue to the result of collecting an HTTP quoted string from input, given position and true.

       2. Collect a sequence of code points that are not U+003B (;) from input, given position.

          Given text/html;charset="shift_jis"iso-2022-jp you end up with text/html;charset=shift_jis.

    9. Otherwise:

       1. Set parameterValue to the result of collecting a sequence of code points that are not U+003B (;) from input, given position.

       2. Remove any trailing HTTP whitespace from parameterValue.

       3. If parameterValue is the empty string, then continue.

    10. If all of the following are true

        • parameterName is not the empty string
        • parameterName solely contains HTTP token code points
        • parameterValue solely contains HTTP quoted-string token code points
        • mimeType’s parameters[parameterName] does not exist

        then set mimeType’s parameters[parameterName] to parameterValue.

12. Return mimeType.

To parse a MIME type from bytes, given a byte sequence input, run these steps:

1. Let string be input, isomorphic decoded.

2. Return the result of parse a MIME type with string.

4.5. Serializing a MIME type

To serialize a MIME type, given a MIME type mimeType, run these steps:

1. Let serialization be the concatenation of mimeType’s type, U+002F (/), and mimeType’s subtype.

2. For each name → value of mimeType’s parameters:

   1. Append U+003B (;) to serialization.

   2. Append name to serialization.

   3. Append U+003D (=) to serialization.

   4. If value does not solely contain HTTP token code points or value is the empty string, then:

      1. Precede each occurrence of U+0022 (") or U+005C (\) in value with U+005C (\).

      2. Prepend U+0022 (") to value.

      3. Append U+0022 (") to value.

   5. Append value to serialization.

3. Return serialization.

To serialize a MIME type to bytes, given a MIME type mimeType, run these steps:

1. Let stringSerialization be the result of serialize a MIME type with mimeType.

2. Return stringSerialization, isomorphic encoded.

4.6. MIME type groups

An image MIME type is a MIME type whose type is "image".

An audio or video MIME type is any MIME type whose type is "audio" or "video", or whose essence is "application/ogg".

A font MIME type is any MIME type whose type is "font", or whose essence is one of the following: [RFC8081]

• application/font-cff
• application/font-off
• application/font-sfnt
• application/font-ttf
• application/font-woff
• application/vnd.ms-fontobject
• application/vnd.ms-opentype

A ZIP-based MIME type is any MIME type whose subtype ends in "+zip" or whose essence is one of the following:

• application/zip

An archive MIME type is any MIME type whose essence is one of the following:

• application/x-rar-compressed
• application/zip
• application/x-gzip

An XML MIME type is any MIME type whose subtype ends in "+xml" or whose essence is "text/xml" or "application/xml". [RFC7303]

An HTML MIME type is any MIME type whose essence is "text/html".

A scriptable MIME type is an XML MIME type, HTML MIME type, or any MIME type whose essence is "application/pdf".

A JavaScript MIME type is any MIME type whose essence is one of the following:

• application/ecmascript
• application/javascript
• application/x-ecmascript
• application/x-javascript
• text/ecmascript
• text/javascript
• text/javascript1.0
• text/javascript1.1
• text/javascript1.2
• text/javascript1.3
• text/javascript1.4
• text/javascript1.5
• text/jscript
• text/livescript
• text/x-ecmascript
• text/x-javascript

A string is a JavaScript MIME type essence match if it is an ASCII case-insensitive match for one of the JavaScript MIME type essence strings.

This hook is used by the type attribute of script elements. [HTML]

A JSON MIME type is any MIME type whose subtype ends in "+json" or whose essence is "application/json" or "text/json".

5. Handling a resource

A resource is ….

For each resource it handles, the user agent must keep track of the following associated metadata:

• A supplied MIME type, the MIME type determined by the supplied MIME type detection algorithm.
• A check-for-apache-bug flag, which defaults to unset.
• A no-sniff flag, which defaults to set if the user agent does not wish to perform sniffing on the resource and unset otherwise.

  The user agent can choose to use outside information, such as previous experience with a site, to determine whether to opt out of sniffing for a particular resource. The user agent can also choose to opt out of sniffing for all resources. However, opting out of sniffing does not exempt the user agent from using the MIME type sniffing algorithm.

• A computed MIME type, the MIME type determined by the MIME type sniffing algorithm.

5.1. Interpreting the resource metadata

The supplied MIME type of a resource is provided to the user agent by an external source associated with that resource. The method of obtaining this information varies depending upon how the resource is retrieved.

To determine the supplied MIME type of a resource, user agents must use the following supplied MIME type detection algorithm:

1. Let supplied-type be null.
2. If the resource is retrieved via HTTP, execute the following steps:
   1. If one or more Content-Type headers are associated with the resource, execute the following steps:
      1. Set supplied-type to the value of the last Content-Type header associated with the resource.

         File extensions are not used to determine the supplied MIME type of a resource retrieved via HTTP because they are unreliable and easily spoofed.

      2. Set the check-for-apache-bug flag if supplied-type is exactly equal to one of the values in the following table:

         Bytes in Hexadecimal	Bytes in ASCII
         74 65 78 74 2F 70 6C 61 69 6E	text/plain
         74 65 78 74 2F 70 6C 61 69 6E 3B 20 63 68 61 72 73 65 74 3D 49 53 4F 2D 38 38 35 39 2D 31	text/plain; charset=ISO-8859-1
         74 65 78 74 2F 70 6C 61 69 6E 3B 20 63 68 61 72 73 65 74 3D 69 73 6F 2D 38 38 35 39 2D 31	text/plain; charset=iso-8859-1
         74 65 78 74 2F 70 6C 61 69 6E 3B 20 63 68 61 72 73 65 74 3D 55 54 46 2D 38	text/plain; charset=UTF-8

         The supplied MIME type detection algorithm detects these exact byte sequences because some older installations of Apache contain a bug that causes them to supply one of these Content-Type headers when serving files with unrecognized MIME types.

   [HTTP-1.1]

3. If the resource is retrieved directly from the file system, set supplied-type to the MIME type provided by the file system.
4. If the resource is retrieved via another protocol (such as FTP), set supplied-type to the MIME type as determined by that protocol, if any.

   [FTP]

5. If supplied-type is not a MIME type, the supplied MIME type is undefined.

   Abort these steps.

6. The supplied MIME type is supplied-type.

5.2. Reading the resource header

A resource header is the byte sequence at the beginning of a resource, as determined by reading the resource header.

To read the resource header, perform the following steps:

1. Let buffer be a byte sequence.
2. Read bytes of the resource into buffer until one of the following conditions is met:
   • the end of the resource is reached.
   • the number of bytes in buffer is greater than or equal to 1445.
   • a reasonable amount of time has elapsed, as determined by the user agent.

   If the number of bytes in buffer is greater than or equal to 1445, the MIME type sniffing algorithm will be deterministic for the majority of cases.

   However, certain factors (such as a slow connection) may prevent the user agent from reading 1445 bytes in a reasonable amount of time.

3. The resource header is buffer.

The resource header need only be determined once per resource.

6. Matching a MIME type pattern

A byte pattern is a byte sequence used as a template to be matched against in the pattern matching algorithm.

A pattern mask is a byte sequence used to determine the significance of bytes being compared against a byte pattern in the pattern matching algorithm.

In a pattern mask, 0xFF indicates the byte is strictly significant, 0xDF indicates that the byte is significant in an ASCII case-insensitive way, and 0x00 indicates that the byte is not significant.

To determine whether a byte sequence matches a particular byte pattern, use the following pattern matching algorithm. It is given a byte sequence input, a byte pattern pattern, a pattern mask mask, and a set of bytes to be ignored ignored, and returns true or false.

1. Assert: pattern’s length is equal to mask’s length.

2. If input’s length is less than pattern’s length, return false.

3. Let s be 0.

4. While s < input’s length:

   1. If ignored does not contain input[s], break.

   2. Set s to s + 1.

5. Let p be 0.

6. While p < pattern’s length:

   1. Let maskedData be the result of applying the bitwise AND operator to input[s] and mask[p].

   2. If maskedData is not equal to pattern[p], return false.

   3. Set s to s + 1.

   4. Set p to p + 1.

7. Return true.

6.1. Matching an image type pattern

To determine which image MIME type byte pattern a byte sequence input matches, if any, use the following image type pattern matching algorithm:

1. Execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given input, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Image MIME Type	Note
   00 00 01 00	FF FF FF FF	None.	image/x-icon	A Windows Icon signature.
   00 00 02 00	FF FF FF FF	None.	image/x-icon	A Windows Cursor signature.
   42 4D	FF FF	None.	image/bmp	The string "BM", a BMP signature.
   47 49 46 38 37 61	FF FF FF FF FF FF	None.	image/gif	The string "GIF87a", a GIF signature.
   47 49 46 38 39 61	FF FF FF FF FF FF	None.	image/gif	The string "GIF89a", a GIF signature.
   52 49 46 46 00 00 00 00 57 45 42 50 56 50	FF FF FF FF 00 00 00 00 FF FF FF FF FF FF	None.	image/webp	The string "RIFF" followed by four bytes followed by the string "WEBPVP".
   89 50 4E 47 0D 0A 1A 0A	FF FF FF FF FF FF FF FF	None.	image/png	An error-checking byte followed by the string "PNG" followed by CR LF SUB LF, the PNG signature.
   FF D8 FF	FF FF FF	None.	image/jpeg	The JPEG Start of Image marker followed by the indicator byte of another marker.

2. Return undefined.

6.2. Matching an audio or video type pattern

To determine which audio or video MIME type byte pattern a byte sequence input matches, if any, use the following audio or video type pattern matching algorithm:

1. Execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given input, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Audio or Video MIME Type	Note
   46 4F 52 4D 00 00 00 00 41 49 46 46	FF FF FF FF 00 00 00 00 FF FF FF FF	None.	audio/aiff	The string "FORM" followed by four bytes followed by the string "AIFF", the AIFF signature.
   49 44 33	FF FF FF	None.	audio/mpeg	The string "ID3", the ID3v2-tagged MP3 signature.
   4F 67 67 53 00	FF FF FF FF FF	None.	application/ogg	The string "OggS" followed by NUL, the Ogg container signature.
   4D 54 68 64 00 00 00 06	FF FF FF FF FF FF FF FF	None.	audio/midi	The string "MThd" followed by four bytes representing the number 6 in 32 bits (big-endian), the MIDI signature.
   52 49 46 46 00 00 00 00 41 56 49 20	FF FF FF FF 00 00 00 00 FF FF FF FF	None.	video/avi	The string "RIFF" followed by four bytes followed by the string "AVI ", the AVI signature.
   52 49 46 46 00 00 00 00 57 41 56 45	FF FF FF FF 00 00 00 00 FF FF FF FF	None.	audio/wave	The string "RIFF" followed by four bytes followed by the string "WAVE", the WAVE signature.

2. If input matches the signature for MP4, return "video/mp4".

3. If input matches the signature for WebM, return "video/webm".

4. If input matches the signature for MP3 without ID3, return "audio/mpeg".

5. Return undefined.

6.2.1. Signature for MP4

To determine whether a byte sequence matches the signature for MP4, use the following steps:

1. Let sequence be the byte sequence to be matched, where sequence[s] is byte s in sequence and sequence[0] is the first byte in sequence.
2. Let length be the number of bytes in sequence.
3. If length is less than 12, return false.
4. Let box-size be the four bytes from sequence[0] to sequence[3], interpreted as a 32-bit unsigned big-endian integer.
5. If length is less than box-size or if box-size modulo 4 is not equal to 0, return false.
6. If the four bytes from sequence[4] to sequence[7] are not equal to 0x66 0x74 0x79 0x70 ("ftyp"), return false.
7. If the three bytes from sequence[8] to sequence[10] are equal to 0x6D 0x70 0x34 ("mp4"), return true.
8. Let bytes-read be 16.

   This ignores the four bytes that correspond to the version number of the "major brand".

9. While bytes-read is less than box-size, continuously loop through these steps:
   1. If the three bytes from sequence[bytes-read] to sequence[bytes-read + 2] are equal to 0x6D 0x70 0x34 ("mp4"), return true.
   2. Increment bytes-read by 4.
10. Return false.

6.2.2. Signature for WebM

To determine whether a byte sequence matches the signature for WebM, use the following steps:

1. Let sequence be the byte sequence to be matched, where sequence[s] is byte s in sequence and sequence[0] is the first byte in sequence.
2. Let length be the number of bytes in sequence.
3. If length is less than 4, return false.
4. If the four bytes from sequence[0] to sequence[3], are not equal to 0x1A 0x45 0xDF 0xA3, return false.
5. Let iter be 4.
6. While iter is less than length and iter is less than 38, continuously loop through these steps:
   1. If the two bytes from sequence[iter] to sequence[iter + 1] are equal to 0x42 0x82,
      1. Increment iter by 2.
      2. If iter is greater or equal than length, abort these steps.
      3. Let number size be the result of parsing a vint starting at sequence[iter].
      4. Increment iter by number size.
      5. If iter is less than length - 4, abort these steps.
      6. Let matched be the result of matching a padded sequence 0x77 0x65 0x62 0x6D ("webm") on sequence at offset iter.
      7. If matched is true, abort these steps and return true.
   2. Increment iter by 1.
7. Return false.

To parse a vint on a byte sequence sequence of size length, starting at index iter use the following steps:

1. Let mask be 128.
2. Let max vint length be 8.
3. Let number size be 1.
4. While number size is less than max vint length, and less than length, continuously loop through these steps:
   1. If the sequence[index] & mask is not zero, abort these steps.
   2. Let mask be the value of mask >> 1.
   3. Increment number size by one.
5. Let index be 0.
6. Let parsed number be sequence[index] & ~mask.
7. Increment index by one.
8. Let bytes remaining be the value of number size.
9. While bytes remaining is not zero, execute there steps:
   1. Let parsed number be parsed number << 8.
   2. Let parsed number be parsed number | sequence[index].
   3. Increment index by one.
   4. If index is greater or equal than length, abort these steps.
   5. Decrement bytes remaining by one.
10. Return parsed number and number size

Matching a padded sequence pattern on a sequence sequence at starting at byte offset and ending at by end means returning true if sequence has a length greater than end, and contains exactly, in the range [offset, end], the bytes in pattern, in the same order, eventually preceded by bytes with a value of 0x00, false otherwise.

6.2.3. Signature for MP3 without ID3

To determine whether a byte sequence matches the signature for MP3 without ID3, use the following steps:

1. Let sequence be the byte sequence to be matched, where sequence[s] is byte s in sequence and sequence[0] is the first byte in sequence.
2. Let length be the number of bytes in sequence.
3. Initialize s to 0.
4. If the result of the operation match mp3 header is false, return false.
5. Parse an mp3 frame on sequence at offset s
6. Let skipped-bytes the return value of the execution of mp3 framesize computation
7. If skipped-bytes is less than 4, or skipped-bytes is greater than s - length, return false.
8. Increment s by skipped-bytes.
9. If the result of the operation match mp3 header operation is false, return false, else, return true.

To match an mp3 header, using a byte sequence sequence of length length at offset s execute these steps:

1. If length is less than 4, return false.
2. If sequence[s] is not equal to 0xff and sequence[s + 1] & 0xe0 is not equal to 0xe0, return false.
3. Let layer be the result of sequence[s + 1] & 0x06 >> 1.
4. If layer is 0, return false.
5. Let bit-rate be sequence[s + 2] & 0xf0 >> 4.
6. If bit-rate is 15, return false.
7. Let sample-rate be sequence[s + 2] & 0x0c >> 2.
8. If sample-rate is 3, return false.
9. Let freq be the value given by sample-rate in the table sample-rate.
10. Let final-layer be the result of 4 - (sequence[s + 1]).
11. If final-layer & 0x06 >> 1 is not 3, return false.
12. Return true.

To compute an mp3 frame size, execute these steps:

1. If version is 1, let scale be 72, else, let scale be 144.
2. Let size be bitrate * scale / freq.
3. If pad is not zero, increment size by 1.
4. Return size.

To parse an mp3 frame, execute these steps:

1. Let version be sequence[s + 1] & 0x18 >> 3.
2. Let bitrate-index be sequence[s + 2] & 0xf0 >> 4.
3. If the version & 0x01 is non-zero, let bitrate be the value given by bitrate-index in the table mp2.5-rates
4. If version & 0x01 is zero, let bitrate be the value given by bitrate-index in the table mp3-rates
5. Let samplerate-index be sequence[s + 2] & 0x0c >> 2.
6. Let samplerate be the value given by samplerate-index in the sample-rate table.
7. Let pad be sequence[s + 2] & 0x02 >> 1.

6.3. Matching a font type pattern

To determine which font MIME type byte pattern a byte sequence input matches, if any, use the following font type pattern matching algorithm:

1. Execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given input, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Font MIME Type	Note
   00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 4C 50	00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 FF FF	None.	application/vnd.ms-fontobject	34 bytes followed by the string "LP", the Embedded OpenType signature.
   00 01 00 00	FF FF FF FF	None.	font/ttf	4 bytes representing the version number 1.0, a TrueType signature.
   4F 54 54 4F	FF FF FF FF	None.	font/otf	The string "OTTO", the OpenType signature.
   74 74 63 66	FF FF FF FF	None.	font/collection	The string "ttcf", the TrueType Collection signature.
   77 4F 46 46	FF FF FF FF	None.	font/woff	The string "wOFF", the Web Open Font Format 1.0 signature.
   77 4F 46 32	FF FF FF FF	None.	font/woff2	The string "wOF2", the Web Open Font Format 2.0 signature.

2. Return undefined.

6.4. Matching an archive type pattern

To determine which archive MIME type byte pattern a byte sequence input matches, if any, use the following archive type pattern matching algorithm:

1. Execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given input, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Archive MIME Type	Note
   1F 8B 08	FF FF FF	None.	application/x-gzip	The GZIP archive signature.
   50 4B 03 04	FF FF FF FF	None.	application/zip	The string "PK" followed by ETX EOT, the ZIP archive signature.
   52 61 72 20 1A 07 00	FF FF FF FF FF FF FF	None.	application/x-rar-compressed	The string "Rar " followed by SUB BEL NUL, the RAR archive signature.

2. Return undefined.

7. Determining the computed MIME type of a resource

To determine the computed MIME type of a resource, user agents must use the following MIME type sniffing algorithm:

1. If the supplied MIME type is undefined or if the supplied MIME type’s essence is "unknown/unknown", "application/unknown", or "*/*", execute the rules for identifying an unknown MIME type with the sniff-scriptable flag equal to the inverse of the no-sniff flag and abort these steps.
2. If the no-sniff flag is set, the computed MIME type is the supplied MIME type.

   Abort these steps.

3. If the check-for-apache-bug flag is set, execute the rules for distinguishing if a resource is text or binary and abort these steps.
4. If the supplied MIME type is an XML MIME type, the computed MIME type is the supplied MIME type.

   Abort these steps.

5. If the supplied MIME type’s essence is "text/html", execute the rules for distinguishing if a resource is a feed or HTML and abort these steps.
6. If the supplied MIME type is an image MIME type supported by the user agent, let matched-type be the result of executing the image type pattern matching algorithm with the resource header as the byte sequence to be matched.
7. If matched-type is not undefined, the computed MIME type is matched-type.

   Abort these steps.

8. If the supplied MIME type is an audio or video MIME type supported by the user agent, let matched-type be the result of executing the audio or video type pattern matching algorithm with the resource header as the byte sequence to be matched.
9. If matched-type is not undefined, the computed MIME type is matched-type.

   Abort these steps.

10. The computed MIME type is the supplied MIME type.

7.1. Identifying a resource with an unknown MIME type

The sniff-scriptable flag is used by the rules for identifying an unknown MIME type to determine whether to sniff for scriptable MIME types.

If the setting of the sniff-scriptable flag is not specified when calling the rules for identifying an unknown MIME type, the sniff-scriptable flag must default to unset.

To determine the computed MIME type of a resource resource with an unknown MIME type, execute the following rules for identifying an unknown MIME type:

1. If the sniff-scriptable flag is set, execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given resource’s resource header, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Computed MIME Type	Note
   3C 21 44 4F 43 54 59 50 45 20 48 54 4D 4C TT	FF FF DF DF DF DF DF DF DF FF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<!DOCTYPE HTML" followed by a tag-terminating byte.
   3C 48 54 4D 4C TT	FF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<HTML" followed by a tag-terminating byte.
   3C 48 45 41 44 TT	FF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<HEAD" followed by a tag-terminating byte.
   3C 53 43 52 49 50 54 TT	FF DF DF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<SCRIPT" followed by a tag-terminating byte.
   3C 49 46 52 41 4D 45 TT	FF DF DF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<IFRAME" followed by a tag-terminating byte.
   3C 48 31 TT	FF DF FF FF	Whitespace bytes.	text/html	The case-insensitive string "<H1" followed by a tag-terminating byte.
   3C 44 49 56 TT	FF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<DIV" followed by a tag-terminating byte.
   3C 46 4F 4E 54 TT	FF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<FONT" followed by a tag-terminating byte.
   3C 54 41 42 4C 45 TT	FF DF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<TABLE" followed by a tag-terminating byte.
   3C 41 TT	FF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<A" followed by a tag-terminating byte.
   3C 53 54 59 4C 45 TT	FF DF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<STYLE" followed by a tag-terminating byte.
   3C 54 49 54 4C 45 TT	FF DF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<TITLE" followed by a tag-terminating byte.
   3C 42 TT	FF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<B" followed by a tag-terminating byte.
   3C 42 4F 44 59 TT	FF DF DF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<BODY" followed by a tag-terminating byte.
   3C 42 52 TT	FF DF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<BR" followed by a tag-terminating byte.
   3C 50 TT	FF DF FF	Whitespace bytes.	text/html	The case-insensitive string "<P" followed by a tag-terminating byte.
   3C 21 2D 2D TT	FF FF FF FF FF	Whitespace bytes.	text/html	The string "<!--" followed by a tag-terminating byte.
   3C 3F 78 6D 6C	FF FF FF FF FF	Whitespace bytes.	text/xml	The string "<?xml".
   25 50 44 46 2D	FF FF FF FF FF	None.	application/pdf	The string "%PDF-", the PDF signature.

   What about feeds?

2. Execute the following steps for each row row in the following table:

   1. Let patternMatched be the result of the pattern matching algorithm given resource’s resource header, the value in the first column of row, the value in the second column of row, and the value in the third column of row.

   2. If patternMatched is true, return the value in the fourth column of row.

   Byte Pattern	Pattern Mask	Leading Bytes to Be Ignored	Computed MIME Type	Note
   25 21 50 53 2D 41 64 6F 62 65 2D	FF FF FF FF FF FF FF FF FF FF FF	None.	application/postscript	The string "%!PS-Adobe-", the PostScript signature.
   FE FF 00 00	FF FF 00 00	None.	text/plain	UTF-16BE BOM
   FF FE 00 00	FF FF 00 00	None.	text/plain	UTF-16LE BOM
   EF BB BF 00	FF FF FF 00	None.	text/plain	UTF-8 BOM

   User agents may implicitly extend this table to support additional MIME types.

   However, user agents should not implicitly extend this table to include additional byte patterns for any computed MIME type already present in this table, as doing so could introduce privilege escalation vulnerabilities.

   User agents must not introduce any privilege escalation vulnerabilities when extending this table.

3. Let matchedType be the result of executing the image type pattern matching algorithm given resource’s resource header.

4. If matchedType is not undefined, return matchedType.

5. Set matchedType to the result of executing the audio or video type pattern matching algorithm given resource’s resource header.

6. If matchedType is not undefined, return matchedType.

7. Set matchedType to the result of executing the archive type pattern matching algorithm given resource’s resource header.

8. If matchedType is not undefined, return matchedType.

9. If resource’s resource header contains no binary data bytes, return "text/plain".

10. Return "application/octet-stream".

7.2. Sniffing a mislabeled binary resource

To determine whether a binary resource has been mislabeled as plain text, execute the following rules for distinguishing if a resource is text or binary:

1. Let length be the number of bytes in the resource header.
2. If length is greater than or equal to 2 and the first 2 bytes of the resource header are equal to 0xFE 0xFF (UTF-16BE BOM) or 0xFF 0xFE (UTF-16LE BOM), the computed MIME type is "text/plain".

   Abort these steps.

3. If length is greater than or equal to 3 and the first 3 bytes of the resource header are equal to 0xEF 0xBB 0xBF (UTF-8 BOM), the computed MIME type is "text/plain".

   Abort these steps.

4. If the resource header contains no binary data bytes, the computed MIME type is "text/plain".

   Abort these steps.

5. The computed MIME type is "application/octet-stream".

   It is critical that the rules for distinguishing if a resource is text or binary never determine the computed MIME type to be a scriptable MIME type, as this could allow a privilege escalation attack.

7.3. Sniffing a mislabeled feed

To determine whether a feed has been mislabeled as HTML, execute the following rules for distinguishing if a resource is a feed or HTML:

1. Let sequence be the resource header, where sequence[s] is byte s in sequence and sequence[0] is the first byte in sequence.
2. Let length be the number of bytes in sequence.
3. Initialize s to 0.
4. If length is greater than or equal to 3 and the three bytes from sequence[0] to sequence[2] are equal to 0xEF 0xBB 0xBF (UTF-8 BOM), increment s by 3.
5. While s is less than length, continuously loop through these steps:
   1. Enter loop L:
      1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

         Abort these steps.

      2. If sequence[s] is equal to 0x3C ("<"), increment s by 1 and exit loop L.
      3. If sequence[s] is not a whitespace byte, the computed MIME type is the supplied MIME type.

         Abort these steps.

      4. Increment s by 1.
   2. Enter loop L:
      1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

         Abort these steps.

      2. If length is greater than or equal to s + 3 and the three bytes from sequence[s] to sequence[s + 2] are equal to 0x21 0x2D 0x2D ("!--"), increment s by 3 and enter loop M:
         1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

            Abort these steps.

         2. If length is greater than or equal to s + 3 and the three bytes from sequence[s] to sequence[s + 2] are equal to 0x2D 0x2D 0x3E ("-->"), increment s by 3 and exit loops M and L.
         3. Increment s by 1.
      3. If length is greater than or equal to s + 1 and sequence[s] is equal to 0x21 ("!"), increment s by 1 and enter loop M:
         1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

            Abort these steps.

         2. If length is greater than or equal to s + 1 and sequence[s] is equal to 0x3E (">"), increment s by 1 and exit loops M and L.
         3. Increment s by 1.
      4. If length is greater than or equal to s + 1 and sequence[s] is equal to 0x3F ("?"), increment s by 1 and enter loop M:
         1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

            Abort these steps.

         2. If length is greater than or equal to s + 2 and the two bytes from sequence[s] to sequence[s + 1] are equal to 0x3F 0x3E ("?>"), increment s by 2 and exit loops M and L.
         3. Increment s by 1.
      5. If length is greater than or equal to s + 3 and the three bytes from sequence[s] to sequence[s + 2] are equal to 0x72 0x73 0x73 ("rss"), the computed MIME type is "application/rss+xml".

         Abort these steps.

      6. If length is greater than or equal to s + 4 and the four bytes from sequence[s] to sequence[s + 3] are equal to 0x66 0x65 0x65 0x64 ("feed"), the computed MIME type is "application/atom+xml".

         Abort these steps.

      7. If length is greater than or equal to s + 7 and the seven bytes from sequence[s] to sequence[s + 6] are equal to 0x72 0x64 0x66 0x3A 0x52 0x44 0x46 ("rdf:RDF"), increment s by 7 and enter loop M:
         1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

            Abort these steps.

         2. If length is greater than or equal to s + 24 and the twenty-four bytes from sequence[s] to sequence[s + 23] are equal to 0x68 0x74 0x74 0x70 0x3A 0x2F 0x2F 0x70 0x75 0x72 0x6C 0x2E 0x6F 0x72 0x67 0x2F 0x72 0x73 0x73 0x2F 0x31 0x2E 0x30 0x2F ("http://purl.org/rss/1.0/"), increment s by 24 and enter loop N:
            1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

               Abort these steps.

            2. If length is greater than or equal to s + 43 and the forty-three bytes from sequence[s] to sequence[s + 42] are equal to 0x68 0x74 0x74 0x70 0x3A 0x2F 0x2F 0x77 0x77 0x77 0x2E 0x77 0x33 0x2E 0x6F 0x72 0x67 0x2F 0x31 0x39 0x39 0x39 0x2F 0x30 0x32 0x2F 0x32 0x32 0x2D 0x72 0x64 0x66 0x2D 0x73 0x79 0x6E 0x74 0x61 0x78 0x2D 0x6E 0x73 0x23 ("http://www.w3.org/1999/02/22-rdf-syntax-ns#"), the computed MIME type is "application/rss+xml".

               Abort these steps.

            3. Increment s by 1.
         3. If length is greater than or equal to s + 24 and the twenty-four bytes from sequence[s] to sequence[s + 23] are equal to 0x68 0x74 0x74 0x70 0x3A 0x2F 0x2F 0x77 0x77 0x77 0x2E 0x77 0x33 0x2E 0x6F 0x72 0x67 0x2F 0x31 0x39 0x39 0x39 0x2F 0x30 0x32 0x2F 0x32 0x32 0x2D 0x72 0x64 0x66 0x2D 0x73 0x79 0x6E 0x74 0x61 0x78 0x2D 0x6E 0x73 0x23 ("http://www.w3.org/1999/02/22-rdf-syntax-ns#"), increment s by 24 and enter loop N:
            1. If sequence[s] is undefined, the computed MIME type is the supplied MIME type.

               Abort these steps.

            2. If length is greater than or equal to s + 43 and the forty-three bytes from sequence[s] to sequence[s + 42] are equal to 0x68 0x74 0x74 0x70 0x3A 0x2F 0x2F 0x70 0x75 0x72 0x6C 0x2E 0x6F 0x72 0x67 0x2F 0x72 0x73 0x73 0x2F 0x31 0x2E 0x30 0x2F ("http://purl.org/rss/1.0/"), the computed MIME type is "application/rss+xml".

               Abort these steps.

            3. Increment s by 1.
         4. Increment s by 1.
      8. The computed MIME type is the supplied MIME type.

         Abort these steps.

6. The computed MIME type is the supplied MIME type.

It might be more efficient for the user agent to implement the rules for distinguishing if a resource is a feed or HTML in parallel with its algorithm for detecting the character encoding of an HTML document.

8. Context-specific sniffing

A context is ….

In certain contexts, it is only useful to identify resources that belong to a certain subset of MIME types.

In such contexts, it is appropriate to use a context-specific sniffing algorithm in place of the MIME type sniffing algorithm in order to determine the computed MIME type of a resource.

A context-specific sniffing algorithm determines the computed MIME type of a resource only if the resource is a MIME type relevant to a particular context.

8.1. Sniffing in a browsing context

Use the MIME type sniffing algorithm.

8.2. Sniffing in an image context

To determine the computed MIME type of a resource with an image MIME type, execute the following rules for sniffing images specifically:

1. If the supplied MIME type is an XML MIME type, the computed MIME type is the supplied MIME type.

   Abort these steps.

2. Let image-type-matched be the result of executing the image type pattern matching algorithm with the resource header as the byte sequence to be matched.
3. If image-type-matched is not undefined, the computed MIME type is image-type-matched.

   Abort these steps.

4. The computed MIME type is the supplied MIME type.

8.3. Sniffing in an audio or video context

To determine the computed MIME type of a resource with an audio or video MIME type, execute the following rules for sniffing audio and video specifically:

1. If the supplied MIME type is an XML MIME type, the computed MIME type is the supplied MIME type.

   Abort these steps.

2. Let audio-or-video-type-matched be the result of executing the audio or video type pattern matching algorithm with the resource header as the byte sequence to be matched.
3. If audio-or-video-type-matched is not undefined, the computed MIME type is audio-or-video-type-matched.

   Abort these steps.

4. The computed MIME type is the supplied MIME type.

8.4. Sniffing in a plugin context

To determine the computed MIME type of a resource fetched in a plugin context, execute the following rules for sniffing in a plugin context:

1. If the supplied MIME type is undefined, the computed MIME type is "application/octet-stream".
2. The computed MIME type is the supplied MIME type.

8.5. Sniffing in a style context

To determine the computed MIME type of a resource fetched in a style context, execute the following rules for sniffing in a style context:

1. If the supplied MIME type is undefined, ….
2. The computed MIME type is the supplied MIME type.

8.6. Sniffing in a script context

To determine the computed MIME type of a resource fetched in a script context, execute the following rules for sniffing in a script context:

1. If the supplied MIME type is undefined, ….
2. The computed MIME type is the supplied MIME type.

8.7. Sniffing in a font context

To determine the computed MIME type of a resource with a font MIME type, execute the following rules for sniffing fonts specifically:

1. If the supplied MIME type is an XML MIME type, the computed MIME type is the supplied MIME type.

   Abort these steps.

2. Let font-type-matched be the result of executing the font type pattern matching algorithm with the resource header as the byte sequence to be matched.
3. If font-type-matched is not undefined, the computed MIME type is font-type-matched.

   Abort these steps.

4. The computed MIME type is the supplied MIME type.

8.8. Sniffing in a text track context

The computed MIME type is "text/vtt".

8.9. Sniffing in a cache manifest context

The computed MIME type is "text/cache-manifest".

Acknowledgments

Special thanks to Adam Barth and Ian Hickson for maintaining previous incarnations of this document.

Thanks also to Alfred Hönes, Andreu Botella, Anne van Kesteren, Boris Zbarsky, Darien Maillet Valentine, David Singer, Domenic Denicola, Henri Sivonen, Jonathan Neal, Joshua Cranmer, Larry Masinter, 罗泽轩, Mariko Kosaka, Mark Pilgrim, Paul Adenot, Peter Occil, Rob Buis, Russ Cox, Simon Pieters, and triple-underscore for their invaluable contributions.

This standard is written by Gordon P. Hemsley (me@gphemsley.org).

Intellectual property rights

Copyright © WHATWG (Apple, Google, Mozilla, Microsoft). This work is licensed under a Creative Commons Attribution 4.0 International License. To the extent portions of it are incorporated into source code, such portions in the source code are licensed under the BSD 3-Clause License instead.

This is the Living Standard. Those interested in the patent-review version should view the Living Standard Review Draft.