]> IMAP Support for UTF-8 Episteme Technology Consulting LLC
503 West Indiana Avenue Urbana IL 61801-4941 United States of America resnick@episteme.net
CNNIC
No.4 South 4th Zhongguancun Street Beijing 100190 China yaojk@cnnic.cn
ICANN
6 Rond Point Schumann, Bd. 1 Brussels 1040 Belgium arnt@gulbrandsen.priv.no
ART extra IMAP This specification extends the Internet Message Access Protocol, specifically IMAP4rev1 (RFC 3501), to support UTF-8 encoded international characters in user names, mail addresses, and message headers. This specification replaces RFC 6855. This specification does not extend IMAP4rev2 (RFC 9051), since that protocol includes everything in this extension.
Introduction This specification forms part of the Email Address Internationalization protocols described in the Email Address Internationalization Framework document . It extends IMAP to permit UTF-8 in headers, as described in "Internationalized Email Headers" . It also adds a mechanism to support mailbox names using the UTF-8 charset. This specification creates two new IMAP capabilities to allow servers to advertise these new extensions. This specification assumes that the IMAP server will be operating in a fully internationalized environment, i.e., one in which all clients accessing the server will be able to accept non-ASCII message header fields and other information, as specified in . At least during a transition period, that assumption will not be realistic for many environments; the issues involved are discussed in below. This specification replaces an earlier, experimental approach to the same problem; see as well as .
Requirements Language The key words "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 BCP 14 when, and only when, they appear in all capitals, as shown here.
"UTF8=ACCEPT" IMAP Capability and UTF-8 in IMAP Quoted-Strings The "UTF8=ACCEPT" capability indicates that the server supports the ability to open mailboxes containing internationalized messages with the "SELECT" and "EXAMINE" commands, and the server can provide UTF-8 responses to the "LIST" and "LSUB" commands. This capability also affects other IMAP extensions that can return mailbox names or their prefixes, such as NAMESPACE and ACL . The "UTF8=ONLY" capability, described in , implies the "UTF8=ACCEPT" capability. A server is said to support "UTF8=ACCEPT" if it advertises either "UTF8=ACCEPT" or "UTF8=ONLY". A client MUST use the "ENABLE" command with the "UTF8=ACCEPT" option (defined in below) to indicate to the server that the client accepts UTF-8 in quoted-strings and supports the "UTF8=ACCEPT" extension. The "ENABLE UTF8=ACCEPT" command is only valid in the authenticated state. The IMAP base specification forbids the use of 8-bit characters in atoms or quoted-strings. Thus, a UTF-8 string can only be sent as a literal. This can be inconvenient from a coding standpoint, and unless the server offers IMAP non-synchronizing literals , this requires an extra round trip for each UTF-8 string sent by the client. When the IMAP server supports "UTF8=ACCEPT", it supports UTF-8 in quoted-strings with the following ABNF syntax : UTF8-3 = UTF8-4 = ]]> When this extended quoting mechanism is used by the client, the server MUST reject, with a "BAD" response, any octet sequences with the high bit set that fail to comply with the formal syntax requirements of UTF-8 . The IMAP server MUST NOT send UTF-8 in quoted-strings to the client unless the client has indicated support for that syntax by using the "ENABLE UTF8=ACCEPT" command. If the server supports "UTF8=ACCEPT", the client MAY use extended quoted syntax with any IMAP argument that permits a string (including astring and nstring). However, if characters outside the US-ASCII repertoire are used in an inappropriate place, the results would be the same as if other syntactically valid but semantically invalid characters were used. Specific cases where UTF-8 characters are permitted or not permitted are described in the following paragraphs. All IMAP servers that support "UTF8=ACCEPT" SHOULD accept UTF-8 in mailbox names, and those that also support the Mailbox International Naming Convention described in , MUST accept UTF-8 in mailbox names and convert them to the appropriate internal format. Mailbox names MUST comply with the Net-Unicode Definition () with the specific exception that they MUST NOT contain control characters (U+0000 - U+001F and U+0080 - U+009F), a delete character (U+007F), a line separator (U+2028), or a paragraph separator (U+2029). Once an IMAP client has enabled UTF-8 support with the "ENABLE UTF8=ACCEPT" command, it MUST NOT issue a "SEARCH" command that contains a charset specification. If an IMAP server receives such a "SEARCH" command in that situation, it SHOULD reject the command with a "BAD" response (due to the conflicting charset labels). This also applies to any IMAP command or extension that includes an optional charset label and associated strings in the command arguments, including the MULTISEARCH extension. For commands with a mandatory charset field, such as SORT and THREAD, servers SHOULD reject charset values other than UTF-8 with a "BAD" response (due to the conflicting charset labels).
"APPEND" Command If the server supports "UTF8=ACCEPT", then the server accepts UTF-8 headers in the "APPEND" command message argument. If an IMAP server supports "UTF8=ACCEPT" and the IMAP client has not issued the "ENABLE UTF8=ACCEPT" command, the server MUST reject, with a "NO" response, an "APPEND" command that includes any 8-bit character in message header fields.
"LOGIN" Command and UTF-8 This specification does not extend the IMAP "LOGIN" command to support UTF-8 usernames and passwords. Whenever a client needs to use UTF-8 usernames or passwords, it MUST use the IMAP "AUTHENTICATE" command, which is already capable of passing UTF-8 usernames and credentials. Although using the IMAP "AUTHENTICATE" command in this way makes it syntactically legal to have a UTF-8 username or password, there is no guarantee that the user provisioning system utilized by the IMAP server will allow such identities. This is an implementation decision and may depend on what identity system the IMAP server is configured to use.
FETCH BODYSTRUCTURE and message/global treats message/global like message/rfc, which means that for some messages, the response to FETCH BODYSTRUCTURE varies depending on whether IMAP4rev1 or IMAP4rev2 is in use. does not extend in this respect. This document extends the media-message ABNF production to match . When IMAP4rev1 and UTF8=ACCEPT has been enabled, the server MAY treat message/global like message/rfc822 when computing the body structure, but MAY also treat it as described in . Clients MUST accept both cases. When IMAP4rev2 and UTF8=ACCEPT are in use, the server MUST behave as described in .
"UTF8=ONLY" Capability The "UTF8=ONLY" capability indicates that the server supports "UTF8=ACCEPT" (see ) and that it requires support for UTF-8 from clients. In particular, this means that the server will send UTF-8 in quoted-strings, and it will not accept the older international mailbox name convention (modified UTF-7 ). Because these are incompatible changes to IMAP, explicit server announcement and client confirmation are necessary: clients MUST use the "ENABLE UTF8=ACCEPT" command before using this server. A server that advertises "UTF8=ONLY" will reject, with a "NO [CANNOT]" response , any command that might require UTF-8 support and is not preceded by an "ENABLE UTF8=ACCEPT" command. IMAP clients that find support for a server that announces "UTF8=ONLY" problematic are encouraged to at least detect the announcement and provide an informative error message to the end user. Because the "UTF8=ONLY" server capability includes support for "UTF8=ACCEPT", the capability string will include, at most, one of those and never both. For the client, "ENABLE UTF8=ACCEPT" is always used -- never "ENABLE UTF8=ONLY".
Dealing with Legacy Clients In most situations, it will be difficult or impossible for the implementer or operator of an IMAP (or POP) server to know whether all of the clients that might access it, or the associated mail store more generally, will be able to support the facilities defined in this document. In almost all cases, servers that conform to this specification will have to be prepared to deal with clients that do not enable the relevant capabilities. Unfortunately, there is no completely satisfactory way to do so other than for systems that wish to receive email that requires SMTPUTF8 capabilities to be sure that all components of those systems -- including IMAP and other clients selected by users -- are upgraded appropriately. When a message that requires SMTPUTF8 is encountered and the client does not enable UTF-8 capability, choices available to the server include hiding the problematic message(s), creating in-band or out-of-band notifications or error messages, or somehow trying to create a surrogate of the message with the intention of providing useful information to that client about what has occurred. Such surrogate messages cannot be actual substitutes for the original message: they will almost always be impossible to reply to (either at all or without loss of information) and the new header fields or specialized constructs for server-client communications may go beyond the requirements of current email specifications (e.g., ). Consequently, such messages may confuse some legacy mail user agents (including IMAP clients) or not provide expected information to users. There are also trade-offs in constructing surrogates of the original message between accepting complexity and additional computation costs in order to try to preserve as much information as possible (for example, in "Post-Delivery Message Downgrading for Internationalized Email Messages" ) and trying to minimize those costs while still providing useful information (for example, in "Simplified POP and IMAP Downgrading for Internationalized Email" ). Implementations that choose to perform downgrading SHOULD use one of the standardized algorithms provided in or . Getting downgrade algorithms right, and minimizing the risk of operational problems and harm to the email system, is tricky and requires careful engineering. These two algorithms are well understood and carefully designed. Because such messages are really surrogates of the original ones, not really "downgraded" ones (although that terminology is often used for convenience), they inevitably have relationships to the originals that the IMAP specification did not anticipate. This brings up two concerns in particular: First, digital signatures computed over and intended for the original message will often not be applicable to the surrogate message, and will often fail signature verification. (It will be possible for some digital signatures to be verified, if they cover only parts of the original message that are not affected in the creation of the surrogate.) Second, servers that may be accessed by the same user with different clients or methods (e.g., POP or webmail systems in addition to IMAP or IMAP clients with different capabilities) will need to exert extreme care to be sure that UIDVALIDITY behaves as the user would expect. Those issues may be especially sensitive if the server caches the surrogate message or computes and stores it when the message arrives with the intent of making either form available depending on client capabilities. Additionally, in order to cope with the case when a server compliant with this extension returns the same UIDVALIDITY to both legacy and "UTF8=ACCEPT"-aware clients, a client upgraded from being non-"UTF8=ACCEPT"-aware MUST discard its cache of messages downloaded from the server. The best (or "least bad") approach for any given environment will depend on local conditions, local assumptions about user behavior, the degree of control the server operator has over client usage and upgrading, the options that are actually available, and so on. It is impossible, at least at the time of publication of this specification, to give good advice that will apply to all situations, or even particular profiles of situations, other than "upgrade legacy clients as soon as possible".
Issues with UTF-8 Header Mailstore When an IMAP server uses a mailbox format that supports UTF-8 headers and it permits selection or examination of that mailbox without issuing "ENABLE UTF8=ACCEPT" first, it is the responsibility of the server to comply with the IMAP base specification and the Internet Message Format with respect to all header information transmitted over the wire. The issue of handling messages containing non-ASCII characters in legacy environments is discussed in .
IANA Considerations the "IMAP Capabilities" registry contained a number of references to . IANA has updated them point to this document instead. The affected references are:
  • UTF8=ACCEPT
  • UTF8=ALL (OBSOLETE)
  • UTF8=APPEND (OBSOLETE)
  • UTF8=ONLY
  • UTF8=USER (OBSOLETE)
Security Considerations The security considerations of UTF-8 and PRECIS Usernames and Passwords apply to this specification, particularly with respect to use of UTF-8 in usernames and passwords. Otherwise, this is not believed to alter the security considerations of IMAP. Special considerations, some of them with security implications, occur if a server that conforms to this specification is accessed by a client that does not, as well as in some more complex situations in which a given message is accessed by multiple clients that might use different protocols and/or support different capabilities. Those issues are discussed in .
References Normative References Informative References
Design Rationale This non-normative section discusses the reasons behind some of the design choices in this specification. The "UTF8=ONLY" mechanism simplifies diagnosis of interoperability problems when legacy support goes away. In the situation where backwards compatibility is not working anyway, the non-conforming "just-send-UTF-8 IMAP" has the advantage that it might work with some legacy clients. However, the difficulty of diagnosing interoperability problems caused by a "just-send-UTF-8 IMAP" mechanism is the reason the "UTF8=ONLY" capability mechanism was chosen.
Changes Since RFC 6855 This non-normative section describes the changes made since .
APPEND UTF8 This document removes APPEND's UTF8 data item, making the UTF8-related syntax compatible with IMAP4rev2 as defined by and making it simpler for clients to support IMAP4rev1 and IMAP4rev2 with the same code. IMAP4rev2 provides roughly the same abilities as but does not include APPEND's UTF8 item. None of , IMAP4rev2, or JMAP specify any way to learn whether a particular message was stored using the UTF8 data item. As of today, an IMAP client cannot learn whether a particular message was stored using the UTF8 data item, nor would it be able to trust that information even if IMAP4rev1 and 2 were extended to provide that information. In July 2023, one of the authors found only one IMAP client that uses the UTF8 data item, and that client uses it incorrectly (it sends the data item for all messages if the server supports UTF8=ACCEPT, without regard to whether a particular message includes any UTF8 at all). For these reasons, it was judged best to revise and adopt the same syntax as IMAP4rev2.
FETCH BODYSTRUCTURE defines a new media type, message/global, which is substantially like message/rfc822 except that the submessage may (also) use the syntax defined in . and define a FETCH item to return the MIME structure of a message, which servers usually compute once and store. None of the RFCs point out to implementers that IMAP4rev1 and IMAP4rev2 are slightly different, so storing the BODYSTRUCTURE in the way servers and clients often do can easily lead to problems. This document makes the syntax optional, making it simple for server authors to implement this extension correctly. This implies that clients need to parse and handle both varieties, which they need to do anyway if they want to support both IMAP4rev1 and IMAP4rev2.
Acknowledgments This document is an almost unchanged copy of , which was written by , , and . Sean has since changed jobs and the current authors do not have a new email address for him. We cannot be sure that he would approve of the changes in this document, so we did not list him as author, but do gratefully acknowledge his work on . replaces him. The next paragraph is a straight copy of the acknowledgments in :
The authors wish to thank the participants of the EAI working group for their contributions to this document, with particular thanks to , , , , , , , , , , , , and for their specific contributions to the discussion.
Many of them also reread the document during this revision.