MAY"> MUST"> MUST NOT"> OPTIONAL"> RECOMMENDED"> REQUIRED"> SHALL"> SHALL NOT"> SHOULD"> SHOULD NOT"> ]> greenbytes GmbH

Hafenweg 16 MuensterNW48155 Germany julian.reschke@greenbytes.de http://greenbytes.de/tech/webdav/

HTTPbis Working Group RFC 2616 defines the Content-Disposition response header field, but points out that it is not part of the HTTP/1.1 Standard. This specification takes over the definition and registration of Content-Disposition, as used in HTTP, and clarifies internationalization aspects. This specification is expected to replace the definition of Content-Disposition in the HTTP/1.1 specification, as currently revised by the IETF HTTPbis working group. See also . Discussion of this draft should take place on the HTTPBIS working group mailing list (ietf-http-wg@w3.org). The current issues list is at and related documents (including fancy diffs) can be found at . The changes in this draft are summarized in .

RFC 2616 defines the Content-Disposition response header field in , but points out that it is not part of the HTTP/1.1 Standard (): Content-Disposition is not part of the HTTP standard, but since it is widely implemented, we are documenting its use and risks for implementers. This specification takes over the definition and registration of Content-Disposition, as used in HTTP. Based on interoperability testing with existing User Agents, it fully defines a profile of the features defined in the Multipurpose Internet Mail Extensions (MIME) variant () of the header field, and also clarifies internationalization aspects. Note: this document does not apply to Content-Disposition header fields appearing in payload bodies transmitted over HTTP, such as when using the media type "multipart/form-data" ().

The 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 . This specification uses the augmented BNF notation defined in , including its rules for implied linear whitespace (LWS).

This specification defines conformance criteria for both senders (usually, HTTP origin servers) and recipients (usually, HTTP user agents) of the Content-Disposition header field. An implementation is considered conformant if it complies with all of the requirements associated with its role. This specification also defines certain forms of the header field-value to be invalid, using both ABNF and prose requirements (), but it does not define special handling of these invalid field-values. Senders &MUST-NOT; generate Content-Disposition header fields that are invalid. Recipients &MAY; take steps to recover a usable field-value from an invalid header field, but &SHOULD-NOT; reject the message outright, unless this is explicitly desirable behaviour (e.g., the implementation is a validator). As such, the default handling of invalid fields is to ignore them.

The Content-Disposition response header field is used to convey additional information about how to process the response payload, and also can be used to attach additional metadata, such as the filename to use when saving the response payload locally.

content-disposition = "Content-Disposition" ":" disposition-type *( ";" disposition-parm ) disposition-type = "inline" | "attachment" | disp-ext-type ; case-insensitive disp-ext-type = token disposition-parm = filename-parm | disp-ext-parm filename-parm = "filename" "=" value | "filename*" "=" ext-value disp-ext-parm = token "=" value | ext-token "=" ext-value ext-token = <the characters in token, followed by "*">

Defined in : token = <token, defined in > quoted-string = <quoted-string, defined in > value = <value, defined in > ; token | quoted-string

Defined in : ext-value = <ext-value, defined in >

Header field values with multiple instances of the same parameter name are invalid. Note that due to the rules for implied linear whitespace (), &OPTIONAL; whitespace can appear between words (token or quoted-string) and separator characters. Furthermore note that the format used for ext-value allows specifying a natural language; this is of limited use for filenames and is likely to be ignored by recipients.

If the disposition type matches "attachment" (case-insensitively), this indicates that the recipient should prompt the user to save the response locally, rather than process it normally (as per its media type). On the other hand, if it matches "inline" (case-insensitively), this implies default processing. Therefore, the disposition type "inline" is only useful when it is augmented with additional parameters, such as the filename (see below). Unknown or unhandled disposition types &SHOULD; be handled by recipients the same way as "attachment" (see also ).

The parameters "filename" and "filename*", to be matched case-insensitively, provide information on how to construct a filename for storing the message payload. Depending on the disposition type, this information might be used right away (in the "save as..." interaction caused for the "attachment" disposition type), or later on (for instance, when the user decides to save the contents of the current page being displayed). The parameters "filename" and "filename*" differ only in that "filename*" uses the encoding defined in , allowing the use of characters not present in the ISO-8859-1 character set (). Many user agent implementations predating this specification do not understand the "filename*" parameter. Therefore, when both "filename" and "filename*" are present in a single header field value, recipients &SHOULD; pick "filename*" and ignore "filename". This way, senders can avoid special-casing specific user agents by sending both the more expressive "filename*" parameter, and the "filename" parameter as fallback for legacy recipients (see for an example). It is essential that recipients treat the specified filename as advisory only, thus be very careful in extracting the desired information. In particular: Recipients &MUST-NOT; be able to write into any location other than one to which they are specifically entitled. To illustrate the problem consider the consequences of being able to overwrite well-known system locations (such as "/etc/passwd"). One strategy to achieve this is to never trust folder name information in the filename parameter, for instance by stripping all but the last path segment and only consider the actual filename (where 'path segment' are the components of the field value delimited by the path separator characters "\" and "/"). Many platforms do not use Internet Media Types () to hold type information in the file system, but rely on filename extensions instead. Trusting the server-provided file extension could introduce a privilege escalation when the saved file is later opened (consider ".exe"). Thus, recipients &SHOULD; ensure that a file extension is used that is safe, optimally matching the media type of the received payload. Recipients &SHOULD; strip or replace character sequences that are known to cause confusion both in user interfaces and in filenames, such as control characters and leading and trailing whitespace. Other aspects recipients need to be aware of are names that have a special meaning in the file system or in shell commands, such as "." and "..", "~", "|", and also device names. Recipients &SHOULD; ignore or substitute names like these. Note: Many user agents do not properly handle the escape character "\" when using the quoted-string form. Furthermore, some user agents erroneously try to perform unescaping of "percent" escapes (see ), and thus might misinterpret filenames containing the percent character followed by two hex digits.

To enable future extensions, recipients &SHOULD; ignore unrecognized parameters (see also ).

Note that defines IANA registries both for disposition types and disposition parameters. This registry is shared by different protocols using Content-Disposition, such as MIME and HTTP. Therefore, not all registered values may make sense in the context of HTTP.

Direct UA to show "save as" dialog, with a filename of "example.html": Content-Disposition: Attachment; filename=example.html

Direct UA to behave as if the Content-Disposition header field wasn't present, but to remember the filename "an example.html" for a subsequent save operation: Content-Disposition: INLINE; FILENAME= "an example.html" Note: this uses the quoted-string form so that the space character can be included.

Direct UA to show "save as" dialog, with a filename containing the Unicode character U+20AC (EURO SIGN): Content-Disposition: attachment; filename*= UTF-8''%e2%82%ac%20rates Here, the encoding defined in is also used to encode the non-ISO-8859-1 character.

Same as above, but adding the "filename" parameter for compatibility with user agents not implementing RFC 5987: Content-Disposition: attachment; filename="EURO rates"; filename*=utf-8''%e2%82%ac%20rates Note: those user agents that do not support the RFC 5987 encoding ignore "filename*" when it occurs after "filename".

The "filename*" parameter (), using the encoding defined in , allows the server to transmit characters outside the ISO-8859-1 character set, and also to optionally specify the language in use. Future parameters might also require internationalization, in which case the same encoding can be used.

Using server-supplied information for constructing local filenames introduces many risks. These are summarized in . Furthermore, implementers also ought to be aware of the Security Considerations applying to HTTP (see ), and also the parameter encoding defined in (see ).

This specification does not introduce any changes to the registration procedures for disposition values and parameters that are defined in .

This document updates the definition of the Content-Disposition HTTP header field in the permanent HTTP header field registry (see ). Content-Disposition http standard IETF this specification ()

Thanks to Adam Barth, Rolf Eike Beer, Stewart Bryant, Bjoern Hoehrmann, Alfred Hoenes, Roar Lauritzsen, Alexey Melnikov, Henrik Nordstrom, and Mark Nottingham for their valuable feedback.

Harvard University

sob@harvard.edu

General keyword University of California, Irvine

fielding@ics.uci.edu

W3C

jg@w3.org

Compaq Computer Corporation

mogul@wrl.dec.com

MIT Laboratory for Computer Science

frystyk@w3.org

Xerox Corporation

masinter@parc.xerox.com

Microsoft Corporation

paulle@microsoft.com

W3C

timbl@w3.org

greenbytes GmbH

Hafenweg 16 MuensterNW48155 Germany julian.reschke@greenbytes.de http://greenbytes.de/tech/webdav/

International Organization for Standardization Innosoft International, Inc.

ned@innosoft.com

First Virtual Holdings

nsb@nsb.fv.com

University of Tennessee

moore@cs.utk.edu

New Century Systems

rens@century.com

QUALCOMM Incorporated

sdorner@qualcomm.com

Department of Computer Science

moore@cs.utk.edu

Innosoft International, Inc.

ned.freed@innosoft.com

University of Tennessee

moore@cs.utk.edu

Xerox Palo Alto Research Center

masinter@parc.xerox.com

Nine by Nine

GK-IETF@ninebynine.org

BEA Systems

mnot@pobox.com

HP Labs

JeffMogul@acm.org

World Wide Web Consortium

timbl@w3.org http://www.w3.org/People/Berners-Lee/

Day Software

fielding@gbiv.com http://roy.gbiv.com/

Adobe Systems Incorporated

LMM@acm.org http://larry.masinter.net/

Compared to , the following normative changes reflecting actual implementations have been made: According to RFC 2616, the disposition type "attachment" only applies to content of type "application/octet-stream". This restriction has been removed, because recipients in practice do not check the content type, and it also discourages properly declaring the media type. RFC 2616 only allows "quoted-string" for the filename parameter. This would be an exceptional parameter syntax, and also doesn't reflect actual use. The definition for the disposition type "inline" () has been re-added with a suggestion for its processing. This specification requires support for the extended parameter encoding defined in .

defines several additional disposition parameters: "creation-date", "modification-date", "quoted-date-time", and "size". The majority of user agents does not implement these, thus they have been omitted from this specification.

By default, HTTP header field parameters cannot carry characters outside the ISO-8859-1 () character encoding (see ). For the "filename" parameter, this of course is an unacceptable restriction. Unfortunately, user agent implementers have not managed to come up with an interoperable approach, although the IETF Standards Track specifies exactly one solution (, clarified and profiled for HTTP in ). For completeness, the sections below describe the various approaches that have been tried, and explains how they are inferior to the RFC 5987 encoding used in this specification.

RFC 2047 defines an encoding mechanism for header fields, but this encoding is not supposed to be used for header field parameters - see : An 'encoded-word' MUST NOT appear within a 'quoted-string'. ... An 'encoded-word' MUST NOT be used in parameter of a MIME Content-Type or Content-Disposition field, or in any structured field body except within a 'comment' or 'phrase'. In practice, some user agents implement the encoding, some do not (exposing the encoded string to the user), and some get confused by it.

Some user agents accept percent encoded () sequences of characters. The character encoding being used for decoding depends on various factors, including the encoding of the referring page, the user agent's locale, its configuration, and also the actual value of the parameter. In practice, this is hard to use because those user agents that do not support it will display the escaped character sequence to the user. For those user agents that do implement this it is difficult to predict what character encoding they actually expect.

Some user agents inspect the value (which defaults to ISO-8859-1 for the quoted-string form) and switch to UTF-8 when it seems to be more likely to be the correct interpretation. As with the approaches above, this is not interoperable and furthermore risks misinterpreting the actual value.

Unfortunately, as of March 2011, neither the encoding defined in RFCs 2231 and 5987, nor any of the alternate approaches discussed above was implemented interoperably. Thus, this specification recommends the approach defined in RFC 5987, which at least has the advantage of actually being specified properly. The table below shows the implementation support for the various approaches: User Agent RFC 2231/5987 RFC 2047 Percent Encoding Encoding Sniffing Chrome yes yes yes yes Firefox yes (*) yes no yes Internet Explorer yes (**) no yes no Konqueror yes no no no Opera yes no no no Safari no no no yes (*) Does not implement the fallback behavior to "filename" described in ; a fix is planned for Firefox 5. (**) Starting with IE9RC, but only implements UTF-8.

To successfully interoperate with existing and future user agents, senders of the Content-Disposition header field are advised to: Include a "filename" parameter when US-ASCII is sufficiently expressive. Use the 'token' form of the filename parameter only when it does not contain disallowed characters (e.g., spaces); in such cases, the quoted-string form should be used. Avoid including the percent character followed by two hexadecimal characters (e.g., %A9) in the filename parameter, since some existing implementations consider it to be an escape character, while others will pass it through unchanged. Avoid including the "\" character in the quoted-string form of the filename parameter, as escaping is not implemented by some user agents, and can be considered as an illegal path character. Avoid using non-ASCII characters in the filename parameter. Although most existing implementations will decode them as ISO-8859-1, some will apply heuristics to detect UTF-8, and thus might fail on certain names. Include a "filename*" parameter where the desired filename cannot be expressed faithfully using the "filename" form. Note that legacy user agents will not process this, and will fall back to using the "filename" parameter's content. When a "filename*" parameter is sent, to also generate a "filename" parameter as a fallback for user agents that do not support the "filename*" form, if possible. This can be done by substituting characters with US-ASCII sequences (e.g., Unicode character point U+00E4 (LATIN SMALL LETTER A WITH DIARESIS) by "ae"). Note that this may not be possible in some locales. When a "filename" parameter is included as a fallback (as per above), "filename" should occur first, due to parsing problems in some existing implementations. Firefox is known to pick the wrong parameter; a bug fix is scheduled for Firefox 5. Use UTF-8 as the encoding of the "filename*" parameter, when present, because at least one existing implementation only implements that encoding. Note that this advice is based upon UA behaviour at the time of writing, and might be superseded. At the time of publication of this document, provides an overview of current levels of support in various implementations.

Note: the issues names in the change log entries for draft-reschke-rfc2183-in-http refer to .

Adjust terminology ("header" -> "header field"). Update rfc2231-in-http reference.

Update rfc2231-in-http reference. Actually define the "filename" parameter. Add internationalization considerations. Add examples using the RFC 5987 encoding. Add overview over other approaches, plus a table reporting implementation status. Add and resolve issue "nodep2183". Add issues "asciivsiso", "deplboth", "quoted", and "registry".

Add and close issue "docfallback". Close issues "asciivsiso", "deplboth", "quoted", and "registry".

Updated to be a Working Draft of the IETF HTTPbis Working Group.

Closed issues: : "handling of unknown disposition types" Slightly updated the notes about the proposed fallback behavior.

Various editorial improvements.

Closed issues: : "state that repeating parameters are invalid" : "warn about %xx in filenames being misinterpreted" : "mention control chars when talking about postprecessing the filename parameter" Update ; Opera 10.63 RC implements the recommended fallback behavior.

Closed issues: : "'modification-date' *is* implemented in Konq 4.5" : "clarify what LWS means for the Content-Disp grammar" : "Avoid passive voice in message requirements" : "text about historical percent-decoding unclear" : "add explanation of language tagging" : "Clarify that C-D spec does not apply to multipart upload"

Updated implementation information (Chrome 9 implements RFC 5987, IE 9 RC implements it for UTF-8 only). Clarify who requirements are on, add a section discussing conformance and handling of invalid field values in general. Closed issues: : "avoid stating ISO-8859-1 default for header param" (the default is still mentioned, but it was clarified what it applies to). : "Path Separator Characters"

Editorial changes: Fixed two typos where the new Conformance section said "Content-Location" instead of "Content-Disposition". Cleaned up terminology ("user agent", "recipient", "sender", "message body", ...). Stated what the escape character for quoted-string is. Explained a use case for "inline" disposition type. Updated implementation notes with respect to the fallback behavior. Added appendix "Advice on Generating Content-Disposition Header Fields".

Closed issues: : "conformance language"

Rephrase the requirement about well-known file system locations, and also clarify that by "last path segment" we mean the actual filename. Added a forward reference from "invalid" to the section that defines a valid header field.