Guidelines for extended attributes

Extended attributes (xattrs) can be set in different namespaces. Linux uses "security", "system", "trusted", and "user". This document only describes and proposes attributes in the "user" namespace.

To avoid future conflicts between attributes, it is recommended that application-specific attributes are namespaced using reverse-DNS naming style. Standard DNS mapping places a subdomain (WWW) before a domain (EXAMPLE) before the top-level (COM), all separated by periods. In Common Extended Attributes, this is done backwards, in which the top level (user) appears first, followed (optionally) by one or more application-specific namespaces, followed by the attribute name, all separated by periods.

For example, if you have written a book, your document might include information about it such as number of words, number of characters, number of chapters, and so on. These attributes might be added by your word processor or you might insert them manually using an attribute editor. Since they would change every time you edited the document, you wouldn't want to include them as a part of your document unless the information was specifically marked as being information about the document ("metadata") in which case either it's kept in the document as a special marked internal value (e.g. an attribute) or is stored outside the document as an attribute. The difference being, at least in theory, one could examine the attribute data about a file without changing the last access date/time of that file. Thus information not directly related to the file itself could be stored externally to the file, such as comments, lists of contributors, and so on.

To prevent one program from misreading attributes written by another - and becoming confused or crashing if what it expects for an attribute is not the same as what was written by the other program - it is recommended that at least one application-specific namespace be used before the attribute, to avoid conflicting use of attributes, unless the attribute is such that it is unique to the file, is unambiguous, and would not be used for any other purpose, e.g. an attribute like "creation-date" or "last-modified-date" might be acceptable as an attribute without a name space, but an attribute such as "Status" might not be as what status represents is probably ambiguous. On the other hand, an attribute such as "document-status" or "document.status" would be a better choice (it would also allow other forms of a "status" value to be associated with the file.)

Then there is the issue of the value for the attribute, if it is not in an application-specific namespace like user.mywordprocessor.lasteditdate as opposed to user.lasteditdate, which value for "lasteditdate" is going to be used? There are probably a half-dozen ways to store dates on computers, many conflicting with each other. For example, is the date 04/05/06 the value April 4, 2006, May 4, 2006, or May 6, 2004? (All three of these interpretations have been validly used for ambiguous date values). And does the date include the time? And is it local time or UTC?

Any attribute for a specific application using the file must be in an application-specific namespace in order to prevent conflicts such as the one given above. The application-specific namespace should probably either be the name of the company producing the application, a general identifier which indicates the application, or both. So, as in the example above, the company making the product "mywordprocessor" would put all of the extended attributes in the form user.mywordprocessor.attribute where "attribute" is whatever extended attribute used by that program. This would prevent some other program accidentally changing or contaminating an attribute it is using.

This recommendation also applies on Mac OS X, where for example the resource fork is stored as an extended attribute with the key "com.apple.?ResourceFork". Attributes specified by freedesktop.org are prefixed with "xdg".

Attribute strings should be in UTF-8 encoding.

General attributes in current use

  • user.mime_type: Sets the MIME type of a file explicitly. This attribute is documented by the "Shared MIME-info Database Specification"(http://standards.freedesktop.org/shared-mime-info-spec). It says: * "An implementation MAY also get a file's MIME type from the user.mime_type extended attribute. The type given here should normally be used in preference to any guessed type, since the user is able to set it explicitly. Applications MAY choose to set the type when saving files. Since many applications and filesystems do not support extended attributes, implementations MUST NOT rely on this method being available."
  • user.charset: The character encoding of a file. This attribute can be useful if the encoding is different than the platform default. It is used by the Apache httpd module mod_mime_xattr (http://0pointer.de/lennart/projects/mod_mime_xattr).
  • user.creator: The name of the application that created the file. The ROX Contact Manager "Contact" sets this value to "Contact" (http://roxos.sunsite.dk/dev-contrib/guido/Contacts/). This meaning of the creator string is different from the meaning in Dublin core

Proposed metadata attributes

Extended attributes are especially useful when they add information that is not present in the in the actual file. Depending on the file origin, an application can add metadata that is not present and/or can not be specified by the file or its format. For example file downloaded from a web server can be tagged with the url, which can be convenient if its source has to be determined in the future. Likewise, an email attachment can when saved be tagged with the message-id of the email. This will make it possible to trace the original message.

These attributes are currently proposed

  • user.xdg.comment: A comment specified by the user. This comment could be displayed by file managers
  • user.xdg.origin.url: Set on a file downloaded from a url. Its value should equal the url it was downloaded from.
  • user.xdg.origin.email.subject: Set on an email attachment when saved to disk. It should get its value from the originating message's "Subject" header
  • user.xdg.origin.email.from: Set on an email attachment when saved to disk. It should get its value from the originating messsage's "From" header. For example '"John Doe" <jdoe@example.com>', or 'jdoe@example.com'
  • user.xdg.origin.email.message-id: Set on an email attachment when saved to disk. It should get its value from the originating message's "Message-Id" header.
  • user.xdg.language: Language of the intellectual content of the resource. The value should follow the syntax described in RFC 3066 in conjunction with ISO 639 language codes. When a file is downloaded using HTTP, the value of the Content-Language HTTP header can if present be copied into this attribute. See also the Language element in Dublin core.
  • user.xdg.creator: Reserved but not yet defined. The string "user" has a different meaning in ROX Contact Manager (creating application) compared with Dublin core (creating person/entity).
  • user.xdg.publisher: Name of the creating application. See also the Publisher element in Dublin core.

Relation to Dublin Core

Dublin Core is a standard for cross-domain information resource description. The Simple Dublin Core Metadata Element Set defines 15 elements. Applications that have knowledge of Dublin Core metadata may choose to set this metadata as extended attributes on a file when it saves it. Some of these elements have counterparts in the xdg name space, but then with more strict or detailed syntax or meaning. Attributes in the Dublin core namespace are currently not further specified.

  • user.dublincore.title
  • user.dublincore.creator See also user.xdg.creator
  • user.dublincore.subject
  • user.dublincore.description
  • user.dublincore.publisher
  • user.dublincore.contributor
  • user.dublincore.date
  • user.dublincore.type
  • user.dublincore.format
  • user.dublincore.identifier
  • user.dublincore.source
  • user.dublincore.language See also user.xdg.language
  • user.dublincore.relation
  • user.dublincore.coverage
  • user.dublincore.rights

Application-specific attributes in current use

  • user.mime_encoding: This attribute defines the MIME encoding to use when serving a file with Apache httpd. It is used by mod_mime_xattr. (Is this attribute useful in general?)
  • user.apache_handler: This attribute is used by mod_mime_xattr to define the Apache handler for a file.
  • user.Beagle.?AttrTime: Used by Beagle
  • user.Beagle.Fingerprint: Used by Beagle
  • user.Beagle.MTime: Used by Beagle
  • user.Beagle.Uid: Used by Beagle

Proposed control attributes

Extended attributes could also act as flags for backup programs, indexing programs and similar, to control their behaviour. Below is an initial proposal for start of discussion, to alter the default behaviour. They are inspired by html attributes to control web crawling.

  • user.xdg.robots.index: On a file: "true" to index, "false" to not index. On a directory: "true" to traverse the into the directory for indexing, "false" for not traversing into it.
  • user.xdg.robots.backup: On a file: "true" to index, "false" to not backup. On a directory: "true" to traverse the into the directory for backup, "false" for not traversing into it.