WOFF File Format 1.0

1. Introduction

This document specifies a simple compressed file format for fonts, designed primarily for use on the Web and known as WOFF (Web Open Font Format). Despite this name, WOFF should be regarded as a container format or "wrapper" for font data in already-existing formats rather than an actual font format in its own right.

The WOFF format is a container for the table-based sfnt structure used in e.g. TrueType [TrueType], OpenType [OpenType] and Open Font Format [OFF] fonts, hereafter referred to as sfnt fonts. A WOFF file is simply a repackaged version of a sfnt font with optional compression of the font data tables. The WOFF file format also allows font metadata and private-use data to be included separately from the font data. WOFF encoding tools convert an input sfnt font into a WOFF formatted file, and user agents restore the sfnt font data for use with a Web document.

The structure and contents of decoded font data exactly match those of a well-formed input font file. Tools producing WOFF files may provide other font editing features such as glyph subsetting, validation or font feature additions but these are considered outside the scope of this format. Independent of these features, both tools and user agents are expected to assure that the validity of the underlying font data is preserved.

Notational Conventions

The all-uppercase key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC-2119]. If these words occur in lower- or mixed case, they should be interpreted in accordance with their normal English meaning.

This document includes sections of text that are called out as "Notes" and set off from the main text of the specification. These notes are intended as informative explanations or clarifications, to serve as hints or guides to implementers and users, but are not part of the normative text.

2. General Requirements

The primary purpose of the WOFF format is to package fonts linked to Web documents by means of CSS @font-face rules. When using such fonts, user agents MUST implement a 'same-origin restriction' on the downloading of WOFF files using the same-origin matching algorithm described in the HTML5 specification ([HTML5 Section 5.3: Origin]). The origin of the stylesheet containing @font-face declarations is not used when deciding whether a WOFF file is same-origin or not, only the origin of containing document is used.

Feature at risk: The WebFonts WG believes that the default Same-Origin restriction would be better applied to all fonts referenced from @font-face, rather than one specific format. Therefore, if CSS3 Fonts [CSS3-Fonts] adds a normative requirement for a Same-Origin restriction, the WebFonts WG will drop it from the WOFF specification and instead refer to CSS3 Fonts.

User agents MUST also implement the ability to relax this restriction using Cross-Origin Resource Sharing [CORS]. Thus, sites can explicitly allow cross-site downloading of WOFF files using the Access-Control-Allow-Origin HTTP header.

Feature at risk: The WebFonts WG suspects that the From-Origin header may be a better way to infer a default Same-Origin for fonts, and the same mechanism can also be used to relax the restriction to allow font sharing across domains. Therefore, once CSS3 Fonts [CSS3-Fonts] mandates a mechanism, WebFonts WG will drop the requirement to use CORS from this specification.

User agents supporting the WOFF file format for linked fonts MUST respect the requirements of the CSS3 Fonts specification ([CSS3-Fonts] Section 4.1: The @font-face rule). In particular, such linked fonts are only available to the documents that reference them; they MUST NOT be made available to other applications or documents on the user's system.

3. Overall File Structure

The structure of WOFF files is similar to the structure of sfnt fonts: a table directory containing lengths and offsets to individual font data tables, followed by the data tables themselves. The sfnt structure is described fully in the TrueType [TrueType], OpenType [OpenType] and Open Font Format [OFF] specifications.

The main body of the file consists of the same collection of font data tables as the input sfnt font, stored in the same order, except that each table MAY be compressed, and the sfnt table directory is replaced by the WOFF table directory.

A complete WOFF file consists of several blocks: a 44-byte header, immediately followed (in this order) by a variable-size table directory, a variable number of font tables, an optional block of extended metadata, and an optional block of private data. Except for padding with a maximum of three null bytes in places where 4-byte alignment of a table length or block offset is specified, there MUST NOT be any extraneous data between the data blocks or font data tables indicated by the WOFF header and table directory, or beyond the last such block or table. If such extraneous data is present a conforming user agent MUST reject the file as invalid. The file MUST also be rejected as invalid if the offsets and lengths of any data blocks or font tables indicate overlapping byte ranges of the file, or ranges that would extend beyond the end of the file.

Data values stored in the WOFF Header and WOFF Table Directory sections are stored in big-endian format, just as values are within sfnt fonts. The following basic data types are used in the description:

All sizes and offsets described in this document are assumed to be in bytes unless otherwise noted.

4. WOFF Header

The header includes an identifying signature and indicates the specific kind of font data included in the file (TrueType or CFF outline data); it also has a font version number, offsets to additional data chunks, and the number of entries in the table directory that immediately follows the header:

The signature field in the WOFF header MUST contain the "magic number" 0x774F4646. If the field does not contain this value, user agents MUST reject the file as invalid.

The flavor field corresponds to the "sfnt version" field found at the beginning of an sfnt file, indicating the type of font data contained. Although only fonts of type 0x00010000 (the version number 1.0 as a 16.16 fixed-point value, indicating TrueType glyph data) and 0x4F54544F (the tag 'OTTO', indicating CFF glyph data) are widely supported at present, it is not an error in the WOFF file if the flavor field contains a different value, indicating a WOFF-packaged version of a different sfnt flavor. (The value 0x74727565 'true' has been used for some TrueType-flavored fonts on Mac OS, for example.) Whether client software will actually support other types of sfnt font data is outside the scope of the WOFF specification, which simply describes how the sfnt is repackaged for Web use.

The WOFF majorVersion and minorVersion fields specify the version number for a given WOFF file, which can be based on the version number of the input font but is not required to be. These fields have no effect on font loading or usage behavior in user agents.

The totalSfntSize field is the sum of the uncompressed font table sizes, each padded to a multiple of 4 bytes, plus the size of the sfnt header and table directory. Thus, this is the size of buffer needed to decode the complete WOFF-packaged font (but not metadata, which is not part of the input sfnt file) into a standard sfnt structure. This value MUST be a multiple of 4, because all font tables including the last are to be padded to a 4-byte boundary. If this value is incorrect, a conforming user agent MUST reject the file as invalid.

The sfnt header includes three fields (searchRange, entrySelector and rangeShift) that are designed to facilitate a binary search of the table directory. As the proper value for each of these fields can be computed directly from the number of font tables, they are not stored in the WOFF file. User agents that decode WOFF files back to an sfnt structure MUST therefore compute the correct values for these fields in the sfnt header, as described in the OpenType/OFF specification.[OFF]

totalSfntSize = 12 // size of sfnt header ("offset table" in the OpenType spec)
totalSfntSize += 16 * numTables // size of table record (directory of font tables)
for each table:
    totalSfntSize += (table.origLength + 3) & 0xFFFFFFFC // table size, padded

If either or both of the metadata or private blocks is not present, the relevant offset and length fields MUST be set to zero. If the metadata or private data offset and length fields indicate byte ranges that overlap other data blocks or tables, or extend beyond the end of the file, a conforming user agent MUST reject the file as invalid.

The header includes a reserved field; this MUST be set to zero. If this field is non-zero, a conforming user agent MUST reject the file as invalid.

5. Table Directory

The table directory is an array of WOFF table directory entries, as defined below. The directory follows immediately after the WOFF file header; therefore, there is no explicit offset in the header pointing to this block. Its size is calculated by multiplying the numTables value in the WOFF header times the size of a single WOFF table directory. Each table directory entry specifies the size and location of a single font data table.

The format of tag values are defined by the specifications for sfnt fonts. The offset and compLength fields identify the location of the compressed font table. The origLength and origCheckSum fields are the length and checksum of the input font table from the table directory of the input font.

The sfnt format specifications require that font tables be aligned on 4-byte boundaries. Font tables whose length is not a multiple of 4 are padded with null bytes up to the next 4-byte boundary. Font data tables in the WOFF file have the same requirement: they MUST begin on 4-byte boundaries and be zero-padded to the next 4-byte boundary. The compLength and origLength fields in the table directory store the exact, unpadded length.

If the offset and compLength values indicate a byte range that overlaps other data blocks or font tables, or if the byte range extends beyond the end of the file, a conforming user agent MUST reject the file as invalid.

If the length of a compressed font table would be the same as or greater than the length of the input font table, the font table MUST be stored uncompressed in the WOFF file and the compLength set equal to the origLength. Tools MAY also opt to leave other tables uncompressed (e.g. all tables less than a certain size), in which case compLength will be equal to origLength for these tables as well. WOFF files containing table directory entries for which compLength is greater than origLength are considered invalid and MUST NOT be loaded by user agents. Files containing compressed font tables that decompress to a size other than origLength are also considered invalid and MUST NOT be loaded.

The sfnt font specifications require that table directory entries are sorted in ascending order of tag value. To simplify processing, WOFF-producing tools MUST produce a table directory with entries in ascending tag value order. User agents MUST likewise assure that the sfnt table directory is recreated in ascending tag order when restoring the font data to its uncompressed state. The ordering of the font tables themselves is independent of the order of directory entries, as described below.

sfnt fonts store a checksum for each table in the table directory, and an overall checksum for the entire font in the head table (see the TrueType, OpenType or Open Font Format specifications for the definition of each calculation). Tools producing WOFF files MUST validate these checksums, and reject the font if errors are found.

A WOFF file contains the same set of font tables as the input font from which it was created. This means that the overall font checksum of a font decompressed from a conformant WOFF file will always match the checksum in the well-formed input font. In the case where the input font included unreferenced data between or after the actual tables, this would affect the overall checksum of the input font, but would be dropped during creation of the WOFF file.

A well-formed input font does not have structural anomalies such as incorrect padding, overlapping font tables, extraneous data between tables (which will be discarded by the WOFF generator), or incorrect checksums.

To ensure that lossless round-trip conversion from sfnt to WOFF and back will be possible, a well-formed input font should conform to certain norms that are not strictly required by the OpenType/OFF specification, although they are common practice:

Font table padding

The OpenType/OFF specification is not entirely clear about whether all tables in an sfnt font must be padded with 0-3 zero bytes to a multiple of 4 bytes in length, or whether this applies only between tables, and the final table of the file may be left unpadded. Most current tools and fonts seem to expect all tables to be padded to a 4-byte boundary, including the last. The WOFF specification assumes this behavior, and specifies that the totalSfntLength field in the WOFF header provides for such padding. To ensure that a given font can be packaged as a WOFF file and then decoded to its original format and give a byte-for-byte identical result, the input font should therefore be padded to a multiple of 4 bytes in length.

No "hidden" data

The OpenType/OFF specification does not explicitly prohibit the presence of "extra" data or padding in between the font tables; as the table directory includes the offset and length of each actual table, such data would simply be ignored. However, the WOFF format makes no provision to preserve such non-font-table data when packaging a font, and therefore it would not survive a round-trip format conversion.

Header fields are correct

The OpenType/OFF header fields (searchRange, entrySelector, rangeShift) that are designed to facilitate a binary search of the table directory must be correct in the input sfnt file, as the WOFF format does not directly preserve these fields but assumes a WOFF decoder will recompute them as needed.

Checksums are correct

The WOFF specification says that table checksums must be validated (and corrected if necessary) by WOFF creators. In order for complete round-trip fidelity, therefore, the checksums in the input sfnt file must also be correct prior to WOFF packaging.

No overlapping tables

The offset and length values in the input sfnt table directory must not indicate overlapping byte ranges of the input font.

The result of creating a WOFF file and then decoding this to regenerate an sfnt font MUST result in a final font that is bitwise-identical to the well-formed input font.. If the input font has defects or anomalies that make this impossible, the WOFF-generating tool SHOULD either reject the font or issue an appropriate warning that lossless round-trip conversion will not be possible.

6. Font Data Tables

The font data tables in the WOFF file are exactly the same as the tables in the input font, except that each table MAY have been compressed. If compressed, it MUST have been compressed by the compress2() function of zlib [Compress2] (or an equivalent, compatible algorithm). User agents use the uncompress() function of zlib [Uncompress](or an equivalent, compatible algorithm) to decompress each table. The underlying format these functions use is described in the ZLIB specification [ZLib]. User agents or other programs that decode WOFF files MUST be able to handle tables that have been compressed. If the decompression function fails for any table, the WOFF file is invalid and MUST NOT be loaded.

The font data tables MUST be stored immediately following the table directory, without gaps except for any padding that is required (up to three null bytes at the end of each table) to ensure 4-byte alignment.

Font tables in WOFF files MUST be stored in the same order as the well-formed input font. The table order is implied by offset values in the table directory; sorting table directory entries into ascending offset value order produces a list of entries in an order equivalent to that of the font tables.

7. Extended Metadata Block

The WOFF file MAY include a block of extended metadata. This may be more extensive and more easily accesible than metadata present in sfnt tables. The metadata block consists of XML data compressed by zlib; the file header specifies both the size of the actual compressed and the original uncompressed size in order to facilitate memory allocation.

The presence (or absence) and content of the metadata block MUST NOT affect font loading or rendering behavior of user agents; it is intended to be purely informative. User agents MAY provide a means for users to view information about fonts (such as a "Font Information" panel). If such information is provided, then they MUST treat the metadata block as the primary source, and MAY fall back to presenting information from the font's name table entries when relevant extended metadata elements are not present.

If present, the metadata MUST be compressed; it is never stored in uncompressed form.

The metadata block MUST follow immediately after the last font table. As all font data tables are padded with up to three null bytes if needed to reach a 4-byte boundary, the beginning of the metadata block will always be 4-byte aligned. If the metadata block is the last block in the WOFF file, there SHOULD be no additional padding after the end of the block.

The extended metadata MUST be well-formed XML encoded in UTF-8.

The schema for the extended metadata XML is described below. If the extended metadata does not match this schema, it is invalid. It is also invalid if it cannot be decompressed by zlib's uncompress() function (or equivalent), or if the length of the decompressed data does not match the metaOrigLength value specified in the WOFF header. Thus, valid metadata is well formed, conforms to the schema below, and is stored in compressed form in the WOFF file. A conforming user agent MUST ignore an invalid metadata block, as if the block were not present.

This description is also available as a RelaxNG schema. In the event of a discrepancy between the RelaxNG schema and the text of the specification, the text takes precedence.

Several elements store their data in text child elements; this is to support localization. The text elements MAY be given a lang attribute in the XML namespace [XML]. For backwards compatibility, lang attributes in the default namepsace are also accepted in older content, and SHOULD be treated the same as xml:lang. New content SHOULD instead use xml:lang.

The syntax of values of the xml:lang attribute can be found in BCP47 [BCP47]. A user agent displaying metadata SHOULD choose a preferred language/locale to display from among those available, following BCP47.

The user agent SHOULD choose which of the available text elements to display as follows:

1. If a text element is available that matches the user's preferred language, as determined via an explicit preference or implied by the current locale, use this language.
2. If the user agent has a concept of a list of "acceptable" languages or defaults, try each of these in turn and use the first one for which a match found.
3. If there is a text element with no xml:lang attribute, use this; in the event that more than one exists, use the first of them.
4. If no match is found yet, use the first text element. (Thus, the metadata creator can determine the "localization of last resort" simply by choosing which language to put first in each group of text elements.)

Such localizable elements are indicated by the statement "This element may be localized using text child elements" in the description below; the internal structure of text elements with xml:lang attributes is not repeated for each element type. In each of these localizable elements, at least one text child element MUST be present, except in the case of the license element (as described below).

metadata element

The main element. This element is REQUIRED.

attributes
version	A version number indicating the format version of the metadata element. This is currently 1.0. This attribute is REQUIRED.
children
uniqueid	Zero or one child elements
vendor	Zero or one child elements
credits	Zero or one child elements
description	Zero or one child elements
license	Zero or one child elements
copyright	Zero or one child elements
trademark	Zero or one child elements
licensee	Zero or one child elements
extension	Zero or more child elements

All first-level child elements of the metadata are OPTIONAL, and may occur in any order as children of the top-level metadata element.

The extension element is intended to allow vendors to include metadata that is not covered by the specific elements defined here, while following a standard model. User agents that provide a means for the user to view WOFF file metadata SHOULD include such extension elements in the metadata presented to the user.

uniqueid element

A unique identifier string for the font. This element is recommended, but not required for the metadata to be valid. This element MUST be a child of the metadata element. This is an empty element.

attributes
id	The identification string. This attribute is REQUIRED.

The string defined in the uniqueid element is not guaranteed to be truly unique, as there is no central registry or authority to ensure this, but it is intended to allow vendors to reliably identify the exact version of a particular font. The use of "reverse-DNS" prefixes to provide a "namespace" is recommended; this can be augmented by additional identification data of the vendor's own design.

The id attribute of the uniqueid element, and of several further metadata elements defined below, is not required to conform to the rules for the XML type ID; its form is at the discretion of the font creator or vendor.

vendor element

Information about the font vendor. This element is recommended, but not required for the metadata to be valid. This element MUST be a child of the metadata element. This is an empty element.

attributes
name	The name of the font vendor. This attribute is REQUIRED.
url	The url for the font vendor. This attribute is OPTIONAL.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

credits element

Credit information for the font. This can include any type of credit the vendor desires: designer, hinter, and so on. This element is OPTIONAL. If present, it MUST be a child of the metadata element and it MUST contain at least one credit element. This element has no attributes.

children
credit	One or more child elements

credit element

A single credit record. If present, it MUST be a child of the credits element. This is an empty element.

attributes
name	The name of the entity being credited. This attribute is REQUIRED.
url	The url for the entity being credited. This attribute is OPTIONAL.
role	The role of the entity being credited. This attribute is OPTIONAL.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

description element

An arbitrary text description of the font's design, its history, etc. This element is OPTIONAL. If present, it MUST be a child of the metadata element. This element may be localized using text child elements.

attributes
url	The url for more information about the font design, history, etc. This attribute is OPTIONAL.
children
text	One or more child elements containing character data and optionally div and/or span children

license element

Licensing information for the font. This element is OPTIONAL. If present, it MUST be a child of the metadata element. This element may be localized using text child elements; however, it is permitted to be empty (for example, if the vendor prefers to just provide a license URL rather than including the actual text of the license.)

attributes
url	The url for the license, more information about the license, etc. This attribute is OPTIONAL.
id	An identifying string for the license. This attribute is OPTIONAL.
children
text	Zero or more child elements containing character data and optionally div and/or span children

copyright element

The copyright for the font. This element is OPTIONAL. If present, it MUST be a child of the metadata element. This element may be localized using text child elements. This element has no attributes.

children
text	One or more child elements containing character data and optionally div and/or span children

trademark element

The trademark for the font. This element is OPTIONAL. If present, it MUST be a child of the metadata element. This element may be localized using text child elements. This element has no attributes.

children
text	One or more child elements containing character data and optionally div and/or span children

licensee element

The licensee of the font. This element is OPTIONAL. If present, it MUST be a child of the metadata element. This is an empty element.

attributes
name	The name of the licensee. This attribute is REQUIRED.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

Although the metadata elements and structure defined above are expected to be sufficient for most needs, an extension mechanism is also provided so that font vendors can include arbitrary metadata items that do not fit the standard elements above:

extension element

A container element for extended metadata provided by the vendor. Zero or more extension elements may be present as children of the top-level metadata element. Each such metadata extension has an optional name, which may be provided in multiple languages, and one or more item elements.

attributes
id	An arbitrary identifier defined by the vendor. This attribute is OPTIONAL.
children
name	Zero or more child elements
item	One or more child elements

item element

At least one item element MUST be present in each extension container.

attributes
id	An arbitrary identifier defined by the vendor. This attribute is OPTIONAL.
children
name	One or more child elements
value	One or more child elements

name element

Zero or more name elements may be used to provide a human-friendly name for the collection of extended metadata items in an extension element. A user agent that displays metadata SHOULD choose the name with most the appropriate language from among those available for each named extension element. This child element is OPTIONAL in extension elements; anonymous extension elements are also permissible.

One or more name elements are also used to provide a human-friendly name for a specific extended metadata item. A user agent that displays metadata SHOULD choose the name with the most appropriate language from among those available for each item element. This child element is REQUIRED in item elements; an item element with no name is invalid and SHOULD be ignored.

attributes
xml:lang	A language tag as defined in BCP47 [BCP47]. This attribute is OPTIONAL.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

value element

One or more value elements are used to provide the value of a specific extended metadata item. A user agent that displays metadata SHOULD choose the value with the most appropriate language from among those available for each item element. This child element is REQUIRED; an item element with no value is invalid and SHOULD be ignored.

attributes
xml:lang	A language tag as defined in BCP47 [BCP47]. This attribute is OPTIONAL.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

Where text elements are used to contain (localizable) content, further structure MAY also be provided using div and span child elements similar to those used in HTML.

text element

An element used to contain a particular localization of a metadata element's contents. This element has a mixed content model; in addition to the child elements mentioned below, it may directly contain character data.

attributes
xml:lang	A language tag (as specified in BCP47 [BCP47]) indicating the language of this particular version of the metadata element's content. This attribute is OPTIONAL; however, for multiple text children of a metadata element to be usefully distinguished, they SHOULD all be tagged with appropriate different language codes.
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.
children
div	Contains a block of text, such as a paragraph or heading.
span	Contains an inline run of text.

div element

A block-level element used, for example, to contain a paragraph.

attributes
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

span element

An inline element used, for example, to indicate a run of text with a different text direction, or in a different language.

attributes
dir	The text direction, either ltr (for "left to right") or rtl (for "right to left"). This attribute is OPTIONAL and, if omitted, defaults to ltr.
class	An arbitrary set of space-separated tokens. This attribute is OPTIONAL.

The text elements used to hold (localizable) text for a number of the individual pieces of metadata thus have a mixed content model consisting of text content, div and span elements; div elements have a mixed content model of text content, div and span elements; and span elements have a mixed content model of text content and span elements. In other words, div can contain other div elements; span can contain other span elements; span does not require a containing div.

Appendix A includes several examples of the content of the metadata block.

8. Private Data Block

The WOFF file MAY include a block of arbitrary data, allowing font creators to include whatever information they wish. The content of this data MUST NOT affect font usage or load behavior of user agents. User agents should make no assumptions about the content of a private block; it may (for example) contain ASCII or Unicode text, or some vendor-defined binary data, and it may be compressed or encrypted, but it has no publicly defined format. Conformant user agents will not assume anything about the structure of this data. Only the font developer or vendor responsible for the private block is expected to understand its contents.

The private data block, if present, MUST be the last block in the WOFF file, following all the font tables and any extended metadata block. The private data block MUST begin on a 4-byte boundary in the WOFF file, with up to three null bytes inserted as padding after any preceding metadata block if needed to ensure this. The end of the private data block MUST correspond to the end of the WOFF file.

Appendix A: Extended Metadata Examples

This appendix is purely informative, not a normative part of the WOFF specification.

This "dummy" metadata block illustrates the use of the metadata elements described in section 7, including the use of multiple text child elements to provide localized versions of certain elements.

<?xml version="1.0" encoding="UTF-8"?>
<metadata version="1.0">
    <uniqueid id="com.example.fontvendor.demofont.rev12345" />
    <vendor name="Font Vendor" url="http://fontvendor.example.com" />
    <credits>
        <credit
            name="Font Designer"
            url="http://fontdesigner.example.com"
            role="Lead" />
        <credit
            name="Another Font Designer"
            url="http://anotherdesigner.example.org"
            role="Contributor" />
        <credit
            name="Yet Another"
            role="Hinting" />
    </credits>
    <description>
        <text xml:lang="en">
            A member of the Demo font family.
            This font is a humanist sans serif style designed
            for optimal legibility in low-resolution environments.
            It can be obtained from fontvendor.example.com.
        </text>
    </description>
    <license url="http://fontvendor.example.com/license"
             id="fontvendor-Web-corporate-v2">
        <text xml:lang="en">A license goes here.</text>
        <text xml:lang="fr">Un permis va ici.</text>
    </license>
    <copyright>
        <text xml:lang="en">Copyright ©2009 Font Vendor"</text>
        <text xml:lang="ko">저작권 ©2009 Font Vendor"</text>
    </copyright>
    <trademark>
        <text xml:lang="en">Demo Font is a trademark of Font Vendor</text>
        <text xml:lang="fr">Demo Font est une marque déposée de Font Vendor</text>
        <text xml:lang="de">Demo Font ist ein eingetragenes Warenzeichen der Font Vendor</text>
        <text xml:lang="ja">Demo FontはFont Vendorの商標である</text>
    </trademark>
    <licensee name="Wonderful Websites, Inc." />
    <extension id="org.example.fonts.metadata.v1">
        <name xml:lang="en">Additional font information</name>
        <name xml:lang="fr">L'information supplémentaire de fonte</name>
        <item id="org.example.fonts.metadata.v1.why">
            <name xml:lang="en">Purpose</name>
            <name xml:lang="fr">But</name>
            <value xml:lang="en">This font exists merely as an example of WOFF packaging.</value>
            <value xml:lang="fr">Cette fonte existe simplement comme exemple de l'empaquetage de WOFF.</value>
        </item>
    </extension>
</metadata>

A real-life example of a simple metadata block (reproduced by permission of FSI FontShop International GmbH).

<?xml version="1.0" encoding="UTF-8"?>
<metadata version="1.0">
   <uniqueid id="com.fontfont.PraterScriptWeb.7.504.1002"/>
   <vendor name="FSI FontShop International GmbH" url="http://www.fontfont.com"/>
   <credits>
       <credit name="Steffen Sauerteig" role="design"/>
       <credit name="Henning Wagenbreth" role="design"/>
       <credit name="FSI FontShop International GmbH" url="http://www.fontfont.com" role="production"/>
   </credits>
   <description>
       <text lang="en">A FontFont for the Web</text>
   </description>
   <license url="http://www.fontfont.com/support/licensing_web.ep" id="fontfont-Web-v1">
       <text lang="en">
	      FontFont Web license v 1.0.
	      For details see http://www.fontfont.com/eula/license_Webfonts_v_1_0.html
	    </text>
   </license>
   <copyright>
       <text lang="en">2009 Henning Wagenbreth, Steffen Sauerteig published by FSI FontShop International GmbH</text>
   </copyright>
   <trademark>
       <text lang="en">Prater is a trademark of FSI FontShop International GmbH</text>
   </trademark>
</metadata>

Another example of a metadata block (reproduced by permission of Ascender Corporation). This is dynamically generated, with the uniqueid and licensee elements modified to be unique for each customer.

<?xml version="1.0" encoding="UTF-8"?>
<metadata version="1.0">
<uniqueid id="037d0f6b-b5e3-b841-05afa509061b" />
<vendor name="Ascender Corp" url="http://www.fontslive.com" />
<license url="http://www.fontslive.com/info/Web-fonts-eula.aspx"
id="ascender-Webfonts-eula-v1">
<text lang="en">This font software is the valuable property of Ascender
Corporation and/or its suppliers and its use by you is covered under the
terms of the Web Font license agreement between you and Ascender
Corporation. You may ONLY use this font software with the licensed Web site.
Except as specifically permitted by the license, you may not copy this font
software. If you have any questions, please contact Ascender Corp.</text>
</license>
<licensee name="FontsLive.com" />
</metadata>

Here is an example of how the div element could be used, modified from a portion of the XML file for Gentium Plus which, when viewed as plain text, has paragraph formatting.

The original metadata file contained:

<description>
 <text lang="en">
	Gentium ("belonging to the nations" in Latin) is a Unicode typeface family 
	designed to enable the many diverse ethnic groups around the world who use 
	the Latin, Cyrillic and Greek scripts to produce readable, high-quality
	publications. The design is intended to be highly readable, reasonably
	compact, and visually attractive. Gentium has won a "Certificate of Excellence
	in Typeface Design" in two major international typeface design competitions: 
	bukva:raz! (2001), TDC2003 (2003).

	The Gentium Plus font family is based on the original design. It currently 
	comes with regular and italic face only, although additional weights are in
	development.

	The goal for this product is to provide a single Unicode-based font family
	that contains a comprehensive inventory of glyphs needed for almost any
	Roman- or Cyrillic-based writing system, whether used for phonetic or
	orthographic needs, and provide a matching Greek face. In addition, there
	is provision for other characters and symbols useful to linguists. This
	font makes use of state-of-the-art font technologies to support complex
	typographic issues, such as the need to position arbitrary combinations
	of base glyphs and diacritics optimally.

   (and so on)
 </text>
</description>

Using the div element, this could become:

<description>
 <text xml:lang="en">
   <div>Gentium ("belonging to the nations" in Latin) is a Unicode
   typeface family designed to enable the many diverse ethnic groups
   around the world who use the Latin, Cyrillic and Greek scripts to
   produce readable, high-quality publications. The design is intended
   to be highly readable, reasonably compact, and visually attractive.
   Gentium has won a "Certificate of Excellence in Typeface
   Design" in two major international typeface design
   competitions: bukva:raz! (2001), TDC2003 (2003).</div>
   <div>The Gentium Plus font family is based on the original design.
   It currently comes with regular and italic face only, although
   additional weights are in development.</div>
   <div>The goal for this product is to provide a single Unicode-based
   font family that contains a comprehensive inventory of glyphs
   needed for almost any Roman- or Cyrillic-based writing system,
   whether used for phonetic or orthographic needs, and provide a
   matching Greek face. In addition, there is provision for other
   characters and symbols useful to linguists. This font makes use of
   state-of-the-art font technologies to support complex typographic
   issues, such as the need to position arbitrary combinations of base
   glyphs and diacritics optimally.</div>

   (and so on)
 </text>
</description>

Appendix B: Media Type registration

This appendix registers a new MIME media type, in conformance with BCP 13 and W3CRegMedia.

Type name:

application

Subtype name:

font-woff

Required parameters:

None.

Optional parameters:

None.

Encoding considerations:

binary.

Security considerations:

Fonts are interpreted data structures that represent collections of glyph outlines, metrics and layout information for various languages and writing systems. Currently, there are many standardized font data tables that allow an unspecified number of entries, and where existing, predefined data fields allow storage of binary data with variable length. There is a significant risk that the flexibility of font data structures may be exploited to hide malicious binary content disguised as a font data component.

WOFF is based on the table-based SFNT (scalable font) format which is highly extensible and offers an opportunity to introduce additional data structures when needed. However, this same extensibility may present specific security concerns – the flexibility and ease of defining new data structures makes it easy for any arbitrary data to be added and hidden inside a font file.

WOFF fonts may contain 'hints' for the alignment of graphical elements of the glyphs with the target display pixel grid, and depending on the font technology utilized in the creation of a font these hints may represent active code interpreted and executed by the font rasterizer. Even though they operate within the confines of the glyph outline conversion system and have no access outside the font rendering machinery, hint instructions can be, however, quite complex, and a maliciously designed complex font could cause undue resource consumption (e.g. memory or CPU cycles) on a machine interpreting it. Indeed, fonts are sufficiently complex that most if not all interpreters cannot be completely protected from malicious fonts without undue performance penalties.

Widespread use of fonts as necessary component of visual content presentation warrants that a careful attention should be given to security considerations whenever a font is either embedded into an electronic document or transmitted alongside media content as a linked resource.

WOFF uses gzip compression. The WOFF header contains the uncompressed length of each compressed table. Applications may therefore constrain the size of memory buffer allocated for decompression and may stop writing if a maliciously crafted WOFF file in fact contains more data than is indicated.

Interoperability considerations:

Published specification:

This media type registration is extracted from the WOFF specification at W3C.

Applications that use this media type:

WOFF is used by Web browsers, often in conjunction with HTML and CSS.

Additional information:

Magic number(s):

The signature field in the WOFF header MUST contain the "magic number" 0x774F4646

File extension(s):

woff

Macintosh file type code(s):

(no code specified)

Macintosh Universal Type Identifier code:

org.w3c.woff

Fragment Identifiers

none.

Person & email address to contact for further information:

Chris Lilley (www-font@w3.org).

Intended usage:

COMMON

Restrictions on usage:

None

Author:

The WOFF specification is a work product of the World Wide Web Consortium's WebFonts Working Group.

Change controller:

The W3C has change control over this specification.

Appendix C: Changes

The following changes have been made, relative to the 27 July 2010 First Public Working Draft. No new features have been added, except for optional enhancements to the metadata format; the remaining changes are clarifications and corrections to existing features.

• Clarified that WOFF is a font packaging or wrapper format, not a new font format as such.
• Clarified padding and byte-alignment requirements throughout
• Preferred the term 'input font' rather than 'original file' to describe the font which is converted to WOFF
• Clarified that any subsetting, checksum-correction, or DSIG invalidating changes prior to WOFF conversion are out of scope for the WOFF specification.
• Clarified meaning of totalSfntSize.
• Defined what is meant by a well formed input font.
• Clarified meaning of 'invalid metadata' - must be well-formed, compressed, and conform to the schema.
• Clarified which elements and attributes in the XML are required; for elements, clarified their required parents.
• Clarified that the id attribute is not of type ID in the XML sense
• Added link to a RelaxNG grammar, which tries to express the same constraints as the prose in a machine-readable manner. The prose is normative in case of any difference.
• Added reminder to check for overflow when decompressing.
• Added a Media Type registration template as required by W3C Media Type registration policies.
• Separated normative and informative references.
• Switched to upper-case for RFC-2119 keywords.
• Added id and class attributes to all testable assertions.
• Added some non-normative explanatory notes.
• Minor clarifications to wording throughout.
• Some re-ordering of text to group related assertions or to improve readability; in particular, the best practices were previously an appendix and are now integrated into the main body of the specification.
• Added this Changes appendix.

The following changes were made as a result of Last Call:

• Language attributes now reference BCP47, and use xml:lang.
• Added optional div and span elements that can be used to provide structure within metadata text.
• Added dir and class attributes to metadata elements.
• Removed summary of conformance requirements (duplicated material from the main body of the specification).
• Somewhat rearranged and clarified description of the metadata format.
• Noted that same-origin requirements are at risk, in the expectation that CSS3 Fonts will handle this instead.
• Mention the need to compute the binary search fields when reconstructing an sfnt header.
• Require that metadata be encoded in UTF-8, rather than allowing either UTF-8 or UTF-16.

A color-coded diff between the editors draft used to prepare the Last Call Working Draft, and the editors draft used to prepare this Candidate Recommendation, is available.

References

Normative References

[BCP47]

BCP 47 Tags for Identifying Languages and Matching of Language Tags

[CORS]

Cross-Origin Resource Sharing, A. van Kesteren, Editor. World Wide Web Consortium, 27 July 2010. The latest version of CORS is available at http://www.w3.org/TR/cors/. (Work in Progress)

[CSS3-Fonts]

CSS Fonts Module Level 3. J. Daggett, Editor. World Wide Web Consortium, 18 June 2009. The latest version of CSS3 Fonts is available at http://www.w3.org/TR/css3-fonts/. (Work in Progress).

[HTML5]

HTML5. I Hickson, Editor. World Wide Web Consortium, 19 October 2010. The latest edition of HTML5 is available at http://www.w3.org/TR/html5/. (Work in Progress).

[OFF]

Open Font Format specification (ISO/IEC 14496-22:2009).

[RFC-2119]

RFC 2119: Key words for use in RFCs to Indicate Requirement Levels. S. Bradner, Editor. Internet Engineering Task Force, March 1997.

[ZLIB]

RFC 1950 ZLIB Compressed Data Format Specification. P. Deutsch, J-L. Gailly, Editors. Internet Engineeering Task Force, May 1996.

Informative References

[Compress2]

zlib compress2() function

[OpenType]

Microsoft OpenType specification, version 1.6. Microsoft, 2009. OpenType is a registered trademark of Microsoft Corporation.

[TrueType]

Apple TrueType Reference manual. Apple, 2002. TrueType is a registered trademark of Apple Computer, Inc.

[Uncompress]

zlib uncompress() function

[XML]

Extensible Markup Language