imapext-2007
view docs/rfc/rfc2683.txt @ 0:ada5e610ab86
imap-2007e
| author | yuuji@gentei.org |
|---|---|
| date | Mon, 14 Sep 2009 15:17:45 +0900 |
| parents | |
| children |
line source
7 Network Working Group B. Leiba
8 Request for Comments: 2683 IBM T.J. Watson Research Center
9 Category: Informational September 1999
12 IMAP4 Implementation Recommendations
14 Status of this Memo
16 This memo provides information for the Internet community. It does
17 not specify an Internet standard of any kind. Distribution of this
18 memo is unlimited.
20 Copyright Notice
22 Copyright (C) The Internet Society (1999). All Rights Reserved.
24 1. Abstract
26 The IMAP4 specification [RFC-2060] describes a rich protocol for use
27 in building clients and servers for storage, retrieval, and
28 manipulation of electronic mail. Because the protocol is so rich and
29 has so many implementation choices, there are often trade-offs that
30 must be made and issues that must be considered when designing such
31 clients and servers. This document attempts to outline these issues
32 and to make recommendations in order to make the end products as
33 interoperable as possible.
35 2. Conventions used in this document
37 In examples, "C:" indicates lines sent by a client that is connected
38 to a server. "S:" indicates lines sent by the server to the client.
40 The words "must", "must not", "should", "should not", and "may" are
41 used with specific meaning in this document; since their meaning is
42 somewhat different from that specified in RFC 2119, we do not put
43 them in all caps here. Their meaning is as follows:
45 must -- This word means that the action described is necessary
46 to ensure interoperability. The recommendation should
47 not be ignored.
48 must not -- This phrase means that the action described will be
49 almost certain to hurt interoperability. The
50 recommendation should not be ignored.
58 Leiba Informational [Page 1]
60 RFC 2683 IMAP4 Implementation Recommendations September 1999
63 should -- This word means that the action described is strongly
64 recommended and will enhance interoperability or
65 usability. The recommendation should not be ignored
66 without careful consideration.
67 should not -- This phrase means that the action described is strongly
68 recommended against, and might hurt interoperability or
69 usability. The recommendation should not be ignored
70 without careful consideration.
71 may -- This word means that the action described is an
72 acceptable implementation choice. No specific
73 recommendation is implied; this word is used to point
74 out a choice that might not be obvious, or to let
75 implementors know what choices have been made by
76 existing implementations.
78 3. Interoperability Issues and Recommendations
80 3.1. Accessibility
82 This section describes the issues related to access to servers and
83 server resources. Concerns here include data sharing and maintenance
84 of client/server connections.
86 3.1.1. Multiple Accesses of the Same Mailbox
88 One strong point of IMAP4 is that, unlike POP3, it allows for
89 multiple simultaneous access to a single mailbox. A user can, thus,
90 read mail from a client at home while the client in the office is
91 still connected; or the help desk staff can all work out of the same
92 inbox, all seeing the same pool of questions. An important point
93 about this capability, though is that NO SERVER IS GUARANTEED TO
94 SUPPORT THIS. If you are selecting an IMAP server and this facility
95 is important to you, be sure that the server you choose to install,
96 in the configuration you choose to use, supports it.
98 If you are designing a client, you must not assume that you can
99 access the same mailbox more than once at a time. That means
101 1. you must handle gracefully the failure of a SELECT command if the
102 server refuses the second SELECT,
103 2. you must handle reasonably the severing of your connection (see
104 "Severed Connections", below) if the server chooses to allow the
105 second SELECT by forcing the first off,
106 3. you must avoid making multiple connections to the same mailbox in
107 your own client (for load balancing or other such reasons), and
108 4. you must avoid using the STATUS command on a mailbox that you have
109 selected (with some server implementations the STATUS command has
110 the same problems with multiple access as do the SELECT and
114 Leiba Informational [Page 2]
116 RFC 2683 IMAP4 Implementation Recommendations September 1999
119 EXAMINE commands).
121 A further note about STATUS: The STATUS command is sometimes used to
122 check a non-selected mailbox for new mail. This mechanism must not
123 be used to check for new mail in the selected mailbox; section 5.2 of
124 [RFC-2060] specifically forbids this in its last paragraph. Further,
125 since STATUS takes a mailbox name it is an independent operation, not
126 operating on the selected mailbox. Because of this, the information
127 it returns is not necessarily in synchronization with the selected
128 mailbox state.
130 3.1.2. Severed Connections
132 The client/server connection may be severed for one of three reasons:
133 the client severs the connection, the server severs the connection,
134 or the connection is severed by outside forces beyond the control of
135 the client and the server (a telephone line drops, for example).
136 Clients and servers must both deal with these situations.
138 When the client wants to sever a connection, it's usually because it
139 has finished the work it needed to do on that connection. The client
140 should send a LOGOUT command, wait for the tagged response, and then
141 close the socket. But note that, while this is what's intended in
142 the protocol design, there isn't universal agreement here. Some
143 contend that sending the LOGOUT and waiting for the two responses
144 (untagged BYE and tagged OK) is wasteful and unnecessary, and that
145 the client can simply close the socket. The server should interpret
146 the closed socket as a log out by the client. The counterargument is
147 that it's useful from the standpoint of cleanup, problem
148 determination, and the like, to have an explicit client log out,
149 because otherwise there is no way for the server to tell the
150 difference between "closed socket because of log out" and "closed
151 socket because communication was disrupted". If there is a
152 client/server interaction problem, a client which routinely
153 terminates a session by breaking the connection without a LOGOUT will
154 make it much more difficult to determine the problem.
156 Because of this disagreement, server designers must be aware that
157 some clients might close the socket without sending a LOGOUT. In any
158 case, whether or not a LOGOUT was sent, the server should not
159 implicitly expunge any messages from the selected mailbox. If a
160 client wants the server to do so, it must send a CLOSE or EXPUNGE
161 command explicitly.
163 When the server wants to sever a connection it's usually due to an
164 inactivity timeout or is because a situation has arisen that has
165 changed the state of the mail store in a way that the server can not
166 communicate to the client. The server should send an untagged BYE
170 Leiba Informational [Page 3]
172 RFC 2683 IMAP4 Implementation Recommendations September 1999
175 response to the client and then close the socket. Sending an
176 untagged BYE response before severing allows the server to send a
177 human-readable explanation of the problem to the client, which the
178 client may then log, display to the user, or both (see section 7.1.5
179 of [RFC-2060]).
181 Regarding inactivity timeouts, there is some controversy. Unlike
182 POP, for which the design is for a client to connect, retrieve mail,
183 and log out, IMAP's design encourages long-lived (and mostly
184 inactive) client/server sessions. As the number of users grows, this
185 can use up a lot of server resources, especially with clients that
186 are designed to maintain sessions for mailboxes that the user has
187 finished accessing. To alleviate this, a server may implement an
188 inactivity timeout, unilaterally closing a session (after first
189 sending an untagged BYE, as noted above). Some server operators have
190 reported dramatic improvements in server performance after doing
191 this. As specified in [RFC-2060], if such a timeout is done it must
192 not be until at least 30 minutes of inactivity. The reason for this
193 specification is to prevent clients from sending commands (such as
194 NOOP) to the server at frequent intervals simply to avert a too-early
195 timeout. If the client knows that the server may not time out the
196 session for at least 30 minutes, then the client need not poll at
197 intervals more frequent than, say, 25 minutes.
199 3.2. Scaling
201 IMAP4 has many features that allow for scalability, as mail stores
202 become larger and more numerous. Large numbers of users, mailboxes,
203 and messages, and very large messages require thought to handle
204 efficiently. This document will not address the administrative
205 issues involved in large numbers of users, but we will look at the
206 other items.
208 3.2.1. Flood Control
210 There are three situations when a client can make a request that will
211 result in a very large response - too large for the client reasonably
212 to deal with: there are a great many mailboxes available, there are a
213 great many messages in the selected mailbox, or there is a very large
214 message part. The danger here is that the end user will be stuck
215 waiting while the server sends (and the client processes) an enormous
216 response. In all of these cases there are things a client can do to
217 reduce that danger.
219 There is also the case where a client can flood a server, by sending
220 an arbitratily long command. We'll discuss that issue, too, in this
221 section.
226 Leiba Informational [Page 4]
228 RFC 2683 IMAP4 Implementation Recommendations September 1999
231 3.2.1.1. Listing Mailboxes
233 Some servers present Usenet newsgroups to IMAP users. Newsgroups,
234 and other such hierarchical mailbox structures, can be very numerous
235 but may have only a few entries at the top level of hierarchy. Also,
236 some servers are built against mail stores that can, unbeknownst to
237 the server, have circular hierarchies - that is, it's possible for
238 "a/b/c/d" to resolve to the same file structure as "a", which would
239 then mean that "a/b/c/d/b" is the same as "a/b", and the hierarchy
240 will never end. The LIST response in this case will be unlimited.
242 Clients that will have trouble with this are those that use
244 C: 001 LIST "" *
246 to determine the mailbox list. Because of this, clients should not
247 use an unqualified "*" that way in the LIST command. A safer
248 approach is to list each level of hierarchy individually, allowing
249 the user to traverse the tree one limb at a time, thus:
251 C: 001 LIST "" %
252 S: * LIST () "/" Banana
253 S: * LIST ...etc...
254 S: 001 OK done
256 and then
258 C: 002 LIST "" Banana/%
259 S: * LIST () "/" Banana/Apple
260 S: * LIST ...etc...
261 S: 002 OK done
263 Using this technique the client's user interface can give the user
264 full flexibility without choking on the voluminous reply to "LIST *".
266 Of course, it is still possible that the reply to
268 C: 005 LIST "" alt.fan.celebrity.%
270 may be thousands of entries long, and there is, unfortunately,
271 nothing the client can do to protect itself from that. This has not
272 yet been a notable problem.
274 Servers that may export circular hierarchies (any server that
275 directly presents a UNIX file system, for instance) should limit the
276 hierarchy depth to prevent unlimited LIST responses. A suggested
277 depth limit is 20 hierarchy levels.
282 Leiba Informational [Page 5]
284 RFC 2683 IMAP4 Implementation Recommendations September 1999
287 3.2.1.2. Fetching the List of Messages
289 When a client selects a mailbox, it is given a count, in the untagged
290 EXISTS response, of the messages in the mailbox. This number can be
291 very large. In such a case it might be unwise to use
293 C: 004 FETCH 1:* ALL
295 to populate the user's view of the mailbox. One good method to avoid
296 problems with this is to batch the requests, thus:
298 C: 004 FETCH 1:50 ALL
299 S: * 1 FETCH ...etc...
300 S: 004 OK done
301 C: 005 FETCH 51:100 ALL
302 S: * 51 FETCH ...etc...
303 S: 005 OK done
304 C: 006 FETCH 101:150 ALL
305 ...etc...
307 Using this method, another command, such as "FETCH 6 BODY[1]" can be
308 inserted as necessary, and the client will not have its access to the
309 server blocked by a storm of FETCH replies. (Such a method could be
310 reversed to fetch the LAST 50 messages first, then the 50 prior to
311 that, and so on.)
313 As a smart extension of this, a well designed client, prepared for
314 very large mailboxes, will not automatically fetch data for all
315 messages AT ALL. Rather, the client will populate the user's view
316 only as the user sees it, possibly pre-fetching selected information,
317 and only fetching other information as the user scrolls to it. For
318 example, to select only those messages beginning with the first
319 unseen one:
321 C: 003 SELECT INBOX
322 S: * 10000 EXISTS
323 S: * 80 RECENT
324 S: * FLAGS (\Answered \Flagged \Deleted \Draft \Seen)
325 S: * OK [UIDVALIDITY 824708485] UID validity status
326 S: * OK [UNSEEN 9921] First unseen message
327 S: 003 OK [READ-WRITE] SELECT completed
328 C: 004 FETCH 9921:* ALL
329 ... etc...
331 If the server does not return an OK [UNSEEN] response, the client may
332 use SEARCH UNSEEN to obtain that value.
338 Leiba Informational [Page 6]
340 RFC 2683 IMAP4 Implementation Recommendations September 1999
343 This mechanism is good as a default presentation method, but only
344 works well if the default message order is acceptable. A client may
345 want to present various sort orders to the user (by subject, by date
346 sent, by sender, and so on) and in that case (lacking a SORT
347 extension on the server side) the client WILL have to retrieve all
348 message descriptors. A client that provides this service should not
349 do it by default and should inform the user of the costs of choosing
350 this option for large mailboxes.
352 3.2.1.3. Fetching a Large Body Part
354 The issue here is similar to the one for a list of messages. In the
355 BODYSTRUCTURE response the client knows the size, in bytes, of the
356 body part it plans to fetch. Suppose this is a 70 MB video clip. The
357 client can use partial fetches to retrieve the body part in pieces,
358 avoiding the problem of an uninterruptible 70 MB literal coming back
359 from the server:
361 C: 022 FETCH 3 BODY[1]<0.20000>
362 S: * 3 FETCH (FLAGS(\Seen) BODY[1]<0> {20000}
363 S: ...data...)
364 S: 022 OK done
365 C: 023 FETCH 3 BODY[1]<20001.20000>
366 S: * 3 FETCH (BODY[1]<20001> {20000}
367 S: ...data...)
368 S: 023 OK done
369 C: 024 FETCH 3 BODY[1]<40001.20000>
370 ...etc...
372 3.2.1.4. BODYSTRUCTURE vs. Entire Messages
374 Because FETCH BODYSTRUCTURE is necessary in order to determine the
375 number of body parts, and, thus, whether a message has "attachments",
376 clients often use FETCH FULL as their normal method of populating the
377 user's view of a mailbox. The benefit is that the client can display
378 a paperclip icon or some such indication along with the normal
379 message summary. However, this comes at a significant cost with some
380 server configurations. The parsing needed to generate the FETCH
381 BODYSTRUCTURE response may be time-consuming compared with that
382 needed for FETCH ENVELOPE. The client developer should consider this
383 issue when deciding whether the ability to add a paperclip icon is
384 worth the tradeoff in performance, especially with large mailboxes.
386 Some clients, rather than using FETCH BODYSTRUCTURE, use FETCH BODY[]
387 (or the equivalent FETCH RFC822) to retrieve the entire message.
388 They then do the MIME parsing in the client. This may give the
389 client slightly more flexibility in some areas (access, for instance,
390 to header fields that aren't returned in the BODYSTRUCTURE and
394 Leiba Informational [Page 7]
396 RFC 2683 IMAP4 Implementation Recommendations September 1999
399 ENVELOPE responses), but it can cause severe performance problems by
400 forcing the transfer of all body parts when the user might only want
401 to see some of them - a user logged on by modem and reading a small
402 text message with a large ZIP file attached may prefer to read the
403 text only and save the ZIP file for later. Therefore, a client
404 should not normally retrieve entire messages and should retrieve
405 message body parts selectively.
407 3.2.1.5. Long Command Lines
409 A client can wind up building a very long command line in an effort to
410 try to be efficient about requesting information from a server. This
411 can typically happen when a client builds a message set from selected
412 messages and doesn't recognise that contiguous blocks of messages may
413 be group in a range. Suppose a user selects all 10,000 messages in a
414 large mailbox and then unselects message 287. The client could build
415 that message set as "1:286,288:10000", but a client that doesn't
416 handle that might try to enumerate each message individually and build
417 "1,2,3,4, [and so on] ,9999,10000". Adding that to the fetch command
418 results in a command line that's almost 49,000 octets long, and,
419 clearly, one can construct a command line that's even longer.
421 A client should limit the length of the command lines it generates to
422 approximately 1000 octets (including all quoted strings but not
423 including literals). If the client is unable to group things into
424 ranges so that the command line is within that length, it should
425 split the request into multiple commands. The client should use
426 literals instead of long quoted strings, in order to keep the command
427 length down.
429 For its part, a server should allow for a command line of at least
430 8000 octets. This provides plenty of leeway for accepting reasonable
431 length commands from clients. The server should send a BAD response
432 to a command that does not end within the server's maximum accepted
433 command length.
435 3.2.2. Subscriptions
437 The client isn't the only entity that can get flooded: the end user,
438 too, may need some flood control. The IMAP4 protocol provides such
439 control in the form of subscriptions. Most servers support the
440 SUBSCRIBE, UNSUBSCRIBE, and LSUB commands, and many users choose to
441 narrow down a large list of available mailboxes by subscribing to the
442 ones that they usually want to see. Clients, with this in mind,
443 should give the user a way to see only subscribed mailboxes. A
444 client that never uses the LSUB command takes a significant usability
445 feature away from the user. Of course, the client would not want to
446 hide the LIST command completely; the user needs to have a way to
450 Leiba Informational [Page 8]
452 RFC 2683 IMAP4 Implementation Recommendations September 1999
455 choose between LIST and LSUB. The usual way to do this is to provide
456 a setting like "show which mailboxes?: [] all [] subscribed only".
458 3.2.3. Searching
460 IMAP SEARCH commands can become particularly troublesome (that is,
461 slow) on mailboxes containing a large number of messages. So let's
462 put a few things in perspective in that regard.
464 The flag searches should be fast. The flag searches (ALL, [UN]SEEN,
465 [UN]ANSWERED, [UN]DELETED, [UN]DRAFT, [UN]FLAGGED, NEW, OLD, RECENT)
466 are known to be used by clients for the client's own use (for
467 instance, some clients use "SEARCH UNSEEN" to find unseen mail and
468 "SEARCH DELETED" to warn the user before expunging messages).
470 Other searches, particularly the text searches (HEADER, TEXT, BODY)
471 are initiated by the user, rather than by the client itself, and
472 somewhat slower performance can be tolerated, since the user is aware
473 that the search is being done (and is probably aware that it might be
474 time-consuming). A smart server might use dynamic indexing to speed
475 commonly used text searches.
477 The client may allow other commands to be sent to the server while a
478 SEARCH is in progress, but at the time of this writing there is
479 little or no server support for parallel processing of multiple
480 commands in the same session (and see "Multiple Accesses of the Same
481 Mailbox" above for a description of the dangers of trying to work
482 around this by doing your SEARCH in another session).
484 Another word about text searches: some servers, built on database
485 back-ends with indexed search capabilities, may return search results
486 that do not match the IMAP spec's "case-insensitive substring"
487 requirements. While these servers are in violation of the protocol,
488 there is little harm in the violation as long as the search results
489 are used only in response to a user's request. Still, developers of
490 such servers should be aware that they ARE violating the protocol,
491 should think carefully about that behaviour, and must be certain that
492 their servers respond accurately to the flag searches for the reasons
493 outlined above.
495 In addition, servers should support CHARSET UTF-8 [UTF-8] in
496 searches.
506 Leiba Informational [Page 9]
508 RFC 2683 IMAP4 Implementation Recommendations September 1999
511 3.3 Avoiding Invalid Requests
513 IMAP4 provides ways for a server to tell a client in advance what is
514 and isn't permitted in some circumstances. Clients should use these
515 features to avoid sending requests that a well designed client would
516 know to be invalid. This section explains this in more detail.
518 3.3.1. The CAPABILITY Command
520 All IMAP4 clients should use the CAPABILITY command to determine what
521 version of IMAP and what optional features a server supports. The
522 client should not send IMAP4rev1 commands and arguments to a server
523 that does not advertize IMAP4rev1 in its CAPABILITY response.
524 Similarly, the client should not send IMAP4 commands that no longer
525 exist in IMAP4rev1 to a server that does not advertize IMAP4 in its
526 CAPABILITY response. An IMAP4rev1 server is NOT required to support
527 obsolete IMAP4 or IMAP2bis commands (though some do; do not let this
528 fact lull you into thinking that it's valid to send such commands to
529 an IMAP4rev1 server).
531 A client should not send commands to probe for the existance of
532 certain extensions. All standard and standards-track extensions
533 include CAPABILITY tokens indicating their presense. All private and
534 experimental extensions should do the same, and clients that take
535 advantage of them should use the CAPABILITY response to determine
536 whether they may be used or not.
538 3.3.2. Don't Do What the Server Says You Can't
540 In many cases, the server, in response to a command, will tell the
541 client something about what can and can't be done with a particular
542 mailbox. The client should pay attention to this information and
543 should not try to do things that it's been told it can't do.
545 Examples:
547 * Do not try to SELECT a mailbox that has the \Noselect flag set.
548 * Do not try to CREATE a sub-mailbox in a mailbox that has the
549 \Noinferiors flag set.
550 * Do not respond to a failing COPY or APPEND command by trying to
551 CREATE the target mailbox if the server does not respond with a
552 [TRYCREATE] response code.
553 * Do not try to expunge a mailbox that has been selected with the
554 [READ-ONLY] response code.
562 Leiba Informational [Page 10]
564 RFC 2683 IMAP4 Implementation Recommendations September 1999
567 3.4. Miscellaneous Protocol Considerations
569 We describe here a number of important protocol-related issues, the
570 misunderstanding of which has caused significant interoperability
571 problems in IMAP4 implementations. One general item is that every
572 implementer should be certain to take note of and to understand
573 section 2.2.2 and the preamble to section 7 of the IMAP4rev1 spec
574 [RFC-2060].
576 3.4.1. Well Formed Protocol
578 We cannot stress enough the importance of adhering strictly to the
579 protocol grammar. The specification of the protocol is quite rigid;
580 do not assume that you can insert blank space for "readability" if
581 none is called for. Keep in mind that there are parsers out there
582 that will crash if there are protocol errors. There are clients that
583 will report every parser burp to the user. And in any case,
584 information that cannot be parsed is information that is lost. Be
585 careful in your protocol generation. And see "A Word About Testing",
586 below.
588 In particular, note that the string in the INTERNALDATE response is
589 NOT an RFC-822 date string - that is, it is not in the same format as
590 the first string in the ENVELOPE response. Since most clients will,
591 in fact, accept an RFC-822 date string in the INTERNALDATE response,
592 it's easy to miss this in your interoperability testing. But it will
593 cause a problem with some client, so be sure to generate the correct
594 string for this field.
596 3.4.2. Special Characters
598 Certain characters, currently the double-quote and the backslash, may
599 not be sent as-is inside a quoted string. These characters must be
600 preceded by the escape character if they are in a quoted string, or
601 else the string must be sent as a literal. Both clients and servers
602 must handle this, both on output (they must send these characters
603 properly) and on input (they must be able to receive escaped
604 characters in quoted strings). Example:
606 C: 001 LIST "" %
607 S: * LIST () "" INBOX
608 S: * LIST () "\\" TEST
609 S: * LIST () "\\" {12}
610 S: "My" mailbox
611 S: 001 OK done
612 C: 002 LIST "" "\"My\" mailbox\\%"
613 S: * LIST () "\\" {17}
614 S: "My" mailbox\Junk
618 Leiba Informational [Page 11]
620 RFC 2683 IMAP4 Implementation Recommendations September 1999
623 S: 002 OK done
625 Note that in the example the server sent the hierarchy delimiter as
626 an escaped character in the quoted string and sent the mailbox name
627 containing imbedded double-quotes as a literal. The client used only
628 quoted strings, escaping both the backslash and the double-quote
629 characters.
631 The CR and LF characters may be sent ONLY in literals; they are not
632 allowed, even if escaped, inside quoted strings.
634 And while we're talking about special characters: the IMAP spec, in
635 the section titled "Mailbox International Naming Convention",
636 describes how to encode mailbox names in modified UTF-7 [UTF-7 and
637 RFC-2060]. Implementations must adhere to this in order to be
638 interoperable in the international market, and servers should
639 validate mailbox names sent by client and reject names that do not
640 conform.
642 As to special characters in userids and passwords: clients must not
643 restrict what a user may type in for a userid or a password. The
644 formal grammar specifies that these are "astrings", and an astring
645 can be a literal. A literal, in turn can contain any 8-bit
646 character, and clients must allow users to enter all 8-bit characters
647 here, and must pass them, unchanged, to the server (being careful to
648 send them as literals when necessary). In particular, some server
649 configurations use "@" in user names, and some clients do not allow
650 that character to be entered; this creates a severe interoperability
651 problem.
653 3.4.3. UIDs and UIDVALIDITY
655 Servers that support existing back-end mail stores often have no good
656 place to save UIDs for messages. Often the existing mail store will
657 not have the concept of UIDs in the sense that IMAP has: strictly
658 increasing, never re-issued, 32-bit integers. Some servers solve
659 this by storing the UIDs in a place that's accessible to end users,
660 allowing for the possibility that the users will delete them. Others
661 solve it by re-assigning UIDs every time a mailbox is selected.
663 The server should maintain UIDs permanently for all messages if it
664 can. If that's not possible, the server must change the UIDVALIDITY
665 value for the mailbox whenever any of the UIDs may have become
666 invalid. Clients must recognize that the UIDVALIDITY has changed and
667 must respond to that condition by throwing away any information that
668 they have saved about UIDs in that mailbox. There have been many
669 problems in this area when clients have failed to do this; in the
670 worst case it will result in loss of mail when a client deletes the
674 Leiba Informational [Page 12]
676 RFC 2683 IMAP4 Implementation Recommendations September 1999
679 wrong piece of mail by using a stale UID.
681 It seems to be a common misunderstanding that "the UIDVALIDITY and
682 the UID, taken together, form a 64-bit identifier that uniquely
683 identifies a message on a server". This is absolutely NOT TRUE.
684 There is no assurance that the UIDVALIDITY values of two mailboxes be
685 different, so the UIDVALIDITY in no way identifies a mailbox. The
686 ONLY purpose of UIDVALIDITY is, as its name indicates, to give the
687 client a way to check the validity of the UIDs it has cached. While
688 it is a valid implementation choice to put these values together to
689 make a 64-bit identifier for the message, the important concept here
690 is that UIDs are not unique between mailboxes; they are only unique
691 WITHIN a given mailbox.
693 Some server implementations have attempted to make UIDs unique across
694 the entire server. This is inadvisable, in that it limits the life
695 of UIDs unnecessarily. The UID is a 32-bit number and will run out
696 in reasonably finite time if it's global across the server. If you
697 assign UIDs sequentially in one mailbox, you will not have to start
698 re-using them until you have had, at one time or another, 2**32
699 different messages in that mailbox. In the global case, you will
700 have to reuse them once you have had, at one time or another, 2**32
701 different messages in the entire mail store. Suppose your server has
702 around 8000 users registered (2**13). That gives an average of 2**19
703 UIDs per user. Suppose each user gets 32 messages (2**5) per day.
704 That gives you 2**14 days (16000+ days = about 45 years) before you
705 run out. That may seem like enough, but multiply the usage just a
706 little (a lot of spam, a lot of mailing list subscriptions, more
707 users) and you limit yourself too much.
709 What's worse is that if you have to wrap the UIDs, and, thus, you
710 have to change UIDVALIDITY and invalidate the UIDs in the mailbox,
711 you have to do it for EVERY mailbox in the system, since they all
712 share the same UID pool. If you assign UIDs per mailbox and you have
713 a problem, you only have to kill the UIDs for that one mailbox.
715 Under extreme circumstances (and this is extreme, indeed), the server
716 may have to invalidate UIDs while a mailbox is in use by a client -
717 that is, the UIDs that the client knows about in its active mailbox
718 are no longer valid. In that case, the server must immediately
719 change the UIDVALIDITY and must communicate this to the client. The
720 server may do this by sending an unsolicited UIDVALIDITY message, in
721 the same form as in response to the SELECT command. Clients must be
722 prepared to handle such a message and the possibly coincident failure
723 of the command in process. For example:
730 Leiba Informational [Page 13]
732 RFC 2683 IMAP4 Implementation Recommendations September 1999
735 C: 032 UID STORE 382 +Flags.silent \Deleted
736 S: * OK [UIDVALIDITY 12345] New UIDVALIDITY value!
737 S: 032 NO UID command rejected because UIDVALIDITY changed!
738 C: ...invalidates local information and re-fetches...
739 C: 033 FETCH 1:* UID
740 ...etc...
742 At the time of the writing of this document, the only server known to
743 do this does so only under the following condition: the client
744 selects INBOX, but there is not yet a physical INBOX file created.
745 Nonetheless, the SELECT succeeds, exporting an empty INBOX with a
746 temporary UIDVALIDITY of 1. While the INBOX remains selected, mail
747 is delivered to the user, which creates the real INBOX file and
748 assigns a permanent UIDVALIDITY (that is likely not to be 1). The
749 server reports the change of UIDVALIDITY, but as there were no
750 messages before, so no UIDs have actually changed, all the client
751 must do is accept the change in UIDVALIDITY.
753 Alternatively, a server may force the client to re-select the
754 mailbox, at which time it will obtain a new UIDVALIDITY value. To do
755 this, the server closes this client session (see "Severed
756 Connections" above) and the client then reconnects and gets back in
757 synch. Clients must be prepared for either of these behaviours.
759 We do not know of, nor do we anticipate the future existance of, a
760 server that changes UIDVALIDITY while there are existing messages,
761 but clients must be prepared to handle this eventuality.
763 3.4.4. FETCH Responses
765 When a client asks for certain information in a FETCH command, the
766 server may return the requested information in any order, not
767 necessarily in the order that it was requested. Further, the server
768 may return the information in separate FETCH responses and may also
769 return information that was not explicitly requested (to reflect to
770 the client changes in the state of the subject message). Some
771 examples:
773 C: 001 FETCH 1 UID FLAGS INTERNALDATE
774 S: * 5 FETCH (FLAGS (\Deleted))
775 S: * 1 FETCH (FLAGS (\Seen) INTERNALDATE "..." UID 345)
776 S: 001 OK done
778 (In this case, the responses are in a different order. Also, the
779 server returned a flag update for message 5, which wasn't part of the
780 client's request.)
786 Leiba Informational [Page 14]
788 RFC 2683 IMAP4 Implementation Recommendations September 1999
791 C: 002 FETCH 2 UID FLAGS INTERNALDATE
792 S: * 2 FETCH (INTERNALDATE "...")
793 S: * 2 FETCH (UID 399)
794 S: * 2 FETCH (FLAGS ())
795 S: 002 OK done
797 (In this case, the responses are in a different order and were
798 returned in separate responses.)
800 C: 003 FETCH 2 BODY[1]
801 S: * 2 FETCH (FLAGS (\Seen) BODY[1] {14}
802 S: Hello world!
803 S: )
804 S: 003 OK done
806 (In this case, the FLAGS response was added by the server, since
807 fetching the body part caused the server to set the \Seen flag.)
809 Because of this characteristic a client must be ready to receive any
810 FETCH response at any time and should use that information to update
811 its local information about the message to which the FETCH response
812 refers. A client must not assume that any FETCH responses will come
813 in any particular order, or even that any will come at all. If after
814 receiving the tagged response for a FETCH command the client finds
815 that it did not get all of the information requested, the client
816 should send a NOOP command to the server to ensure that the server
817 has an opportunity to send any pending EXPUNGE responses to the
818 client (see [RFC-2180]).
820 3.4.5. RFC822.SIZE
822 Some back-end mail stores keep the mail in a canonical form, rather
823 than retaining the original MIME format of the messages. This means
824 that the server must reassemble the message to produce a MIME stream
825 when a client does a fetch such as RFC822 or BODY[], requesting the
826 entire message. It also may mean that the server has no convenient
827 way to know the RFC822.SIZE of the message. Often, such a server
828 will actually have to build the MIME stream to compute the size, only
829 to throw the stream away and report the size to the client.
831 When this is the case, some servers have chosen to estimate the size,
832 rather than to compute it precisely. Such an estimate allows the
833 client to display an approximate size to the user and to use the
834 estimate in flood control considerations (q.v.), but requires that
835 the client not use the size for things such as allocation of buffers,
836 because those buffers might then be too small to hold the actual MIME
837 stream. Instead, a client should use the size that's returned in the
838 literal when you fetch the data.
842 Leiba Informational [Page 15]
844 RFC 2683 IMAP4 Implementation Recommendations September 1999
847 The protocol requires that the RFC822.SIZE value returned by the
848 server be EXACT. Estimating the size is a protocol violation, and
849 server designers must be aware that, despite the performance savings
850 they might realize in using an estimate, this practice will cause
851 some clients to fail in various ways. If possible, the server should
852 compute the RFC822.SIZE for a particular message once, and then save
853 it for later retrieval. If that's not possible, the server must
854 compute the value exactly every time. Incorrect estimates do cause
855 severe interoperability problems with some clients.
857 3.4.6. Expunged Messages
859 If the server allows multiple connections to the same mailbox, it is
860 often possible for messages to be expunged in one client unbeknownst
861 to another client. Since the server is not allowed to tell the
862 client about these expunged messages in response to a FETCH command,
863 the server may have to deal with the issue of how to return
864 information about an expunged message. There was extensive
865 discussion about this issue, and the results of that discussion are
866 summarized in [RFC-2180]. See that reference for a detailed
867 explanation and for recommendations.
869 3.4.7. The Namespace Issue
871 Namespaces are a very muddy area in IMAP4 implementation right now
872 (see [NAMESPACE] for a proposal to clear the water a bit). Until the
873 issue is resolved, the important thing for client developers to
874 understand is that some servers provide access through IMAP to more
875 than just the user's personal mailboxes, and, in fact, the user's
876 personal mailboxes may be "hidden" somewhere in the user's default
877 hierarchy. The client, therefore, should provide a setting wherein
878 the user can specify a prefix to be used when accessing mailboxes. If
879 the user's mailboxes are all in "~/mail/", for instance, then the
880 user can put that string in the prefix. The client would then put
881 the prefix in front of any name pattern in the LIST and LSUB
882 commands:
884 C: 001 LIST "" ~/mail/%
886 (See also "Reference Names in the LIST Command" below.)
888 3.4.8. Creating Special-Use Mailboxes
890 It may seem at first that this is part of the namespace issue; it is
891 not, and is only indirectly related to it. A number of clients like
892 to create special-use mailboxes with particular names. Most
893 commonly, clients with a "trash folder" model of message deletion
894 want to create a mailbox with the name "Trash" or "Deleted". Some
898 Leiba Informational [Page 16]
900 RFC 2683 IMAP4 Implementation Recommendations September 1999
903 clients want to create a "Drafts" mailbox, an "Outbox" mailbox, or a
904 "Sent Mail" mailbox. And so on. There are two major
905 interoperability problems with this practice:
907 1. different clients may use different names for mailboxes with
908 similar functions (such as "Trash" and "Deleted"), or may manage
909 the same mailboxes in different ways, causing problems if a user
910 switches between clients and
911 2. there is no guarantee that the server will allow the creation of
912 the desired mailbox.
914 The client developer is, therefore, well advised to consider
915 carefully the creation of any special-use mailboxes on the server,
916 and, further, the client must not require such mailbox creation -
917 that is, if you do decide to do this, you must handle gracefully the
918 failure of the CREATE command and behave reasonably when your
919 special-use mailboxes do not exist and can not be created.
921 In addition, the client developer should provide a convenient way for
922 the user to select the names for any special-use mailboxes, allowing
923 the user to make these names the same in all clients used and to put
924 them where the user wants them.
926 3.4.9. Reference Names in the LIST Command
928 Many implementers of both clients and servers are confused by the
929 "reference name" on the LIST command. The reference name is intended
930 to be used in much the way a "cd" (change directory) command is used
931 on Unix, PC DOS, Windows, and OS/2 systems. That is, the mailbox
932 name is interpreted in much the same way as a file of that name would
933 be found if one had done a "cd" command into the directory specified
934 by the reference name. For example, in Unix we have the following:
936 > cd /u/jones/junk
937 > vi banana [file is "/u/jones/junk/banana"]
938 > vi stuff/banana [file is "/u/jones/junk/stuff/banana"]
939 > vi /etc/hosts [file is "/etc/hosts"]
941 In the past, there have been several interoperability problems with
942 this. First, while some IMAP servers are built on Unix or PC file
943 systems, many others are not, and the file system semantics do not
944 make sense in those configurations. Second, while some IMAP servers
945 expose the underlying file system to the clients, others allow access
946 only to the user's personal mailboxes, or to some other limited set
947 of files, making such file-system-like semantics less meaningful.
948 Third, because the IMAP spec leaves the interpretation of the
949 reference name as "implementation-dependent", in the past the various
950 server implementations handled it in vastly differing ways.
954 Leiba Informational [Page 17]
956 RFC 2683 IMAP4 Implementation Recommendations September 1999
959 The following recommendations are the result of significant
960 operational experience, and are intended to maximize
961 interoperability.
963 Server implementations must implement the reference argument in a way
964 that matches the intended "change directory" operation as closely as
965 possible. As a minimum implementation, the reference argument may be
966 prepended to the mailbox name (while suppressing double delimiters;
967 see the next paragraph). Even servers that do not provide a way to
968 break out of the current hierarchy (see "breakout facility" below)
969 must provide a reasonable implementation of the reference argument,
970 as described here, so that they will interoperate with clients that
971 use it.
973 Server implementations that prepend the reference argument to the
974 mailbox name should insert a hierarchy delimiter between them, and
975 must not insert a second if one is already present:
977 C: A001 LIST ABC DEF
978 S: * LIST () "/" ABC/DEF <=== should do this
979 S: A001 OK done
981 C: A002 LIST ABC/ /DEF
982 S: * LIST () "/" ABC//DEF <=== must not do this
983 S: A002 OK done
985 On clients, the reference argument is chiefly used to implement a
986 "breakout facility", wherein the user may directly access a mailbox
987 outside the "current directory" hierarchy. Client implementations
988 should have an operational mode that does not use the reference
989 argument. This is to interoperate with older servers that did not
990 implement the reference argument properly. While it's a good idea to
991 give the user access to a breakout facility, clients that do not
992 intend to do so should not use the reference argument at all.
994 Client implementations should always place a trailing hierarchy
995 delimiter on the reference argument. This is because some servers
996 prepend the reference argument to the mailbox name without inserting
997 a hierarchy delimiter, while others do insert a hierarchy delimiter
998 if one is not already present. A client that puts the delimiter in
999 will work with both varieties of server.
1001 Client implementations that implement a breakout facility should
1002 allow the user to choose whether or not to use a leading hierarchy
1003 delimiter on the mailbox argument. This is because the handling of a
1004 leading mailbox hierarchy delimiter also varies from server to
1005 server, and even between different mailstores on the same server. In
1006 some cases, a leading hierarchy delimiter means "discard the
1010 Leiba Informational [Page 18]
1012 RFC 2683 IMAP4 Implementation Recommendations September 1999
1015 reference argument" (implementing the intended breakout facility),
1016 thus:
1018 C: A001 LIST ABC/ /DEF
1019 S: * LIST () "/" /DEF
1020 S: A001 OK done
1022 In other cases, however, the two are catenated and the extra
1023 hierarchy delimiter is discarded, thus:
1025 C: A001 LIST ABC/ /DEF
1026 S: * LIST () "/" ABC/DEF
1027 S: A001 OK done
1029 Client implementations must not assume that the server supports a
1030 breakout facility, but may provide a way for the user to use one if
1031 it is available. Any breakout facility should be exported to the
1032 user interface. Note that there may be other "breakout" characters
1033 besides the hierarchy delimiter (for instance, UNIX filesystem
1034 servers are likely to use a leading "~" as well), and that their
1035 interpretation is server-dependent.
1037 3.4.10. Mailbox Hierarchy Delimiters
1039 The server's selection of what to use as a mailbox hierarchy
1040 delimiter is a difficult one, involving several issues: What
1041 characters do users expect to see? What characters can they enter
1042 for a hierarchy delimiter if it is desired (or required) that the
1043 user enter it? What character can be used for the hierarchy
1044 delimiter, noting that the chosen character can not otherwise be used
1045 in the mailbox name?
1047 Because some interfaces show users the hierarchy delimiters or allow
1048 users to enter qualified mailbox names containing them, server
1049 implementations should use delimiter characters that users generally
1050 expect to see as name separators. The most common characters used
1051 for this are "/" (as in Unix file names), "\" (as in OS/2 and Windows
1052 file names), and "." (as in news groups). There is little to choose
1053 among these apart from what users may expect or what is dictated by
1054 the underlying file system, if any. One consideration about using
1055 "\" is that it's also a special character in the IMAP protocol. While
1056 the use of other hierarchy delimiter characters is permissible, A
1057 DESIGNER IS WELL ADVISED TO STAY WITH ONE FROM THIS SET unless the
1058 server is intended for special purposes only. Implementers might be
1059 thinking about using characters such as "-", "_", ";", "&", "#", "@",
1060 and "!", but they should be aware of the surprise to the user as well
1061 as of the effect on URLs and other external specifications (since
1062 some of these characters have special meanings there). Also, a
1066 Leiba Informational [Page 19]
1068 RFC 2683 IMAP4 Implementation Recommendations September 1999
1071 server that uses "\" (and clients of such a server) must remember to
1072 escape that character in quoted strings or to send literals instead.
1073 Literals are recommended over escaped characters in quoted strings in
1074 order to maintain compatibility with older IMAP versions that did not
1075 allow escaped characters in quoted strings (but check the grammar to
1076 see where literals are allowed):
1078 C: 001 LIST "" {13}
1079 S: + send literal
1080 C: this\%\%\%\h*
1081 S: * LIST () "\\" {27}
1082 S: this\is\a\mailbox\hierarchy
1083 S: 001 OK LIST complete
1085 In any case, a server should not use normal alpha-numeric characters
1086 (such as "X" or "0") as delimiters; a user would be very surprised to
1087 find that "EXPENDITURES" actually represented a two-level hierarchy.
1088 And a server should not use characters that are non-printable or
1089 difficult or impossible to enter on a standard US keyboard. Control
1090 characters, box-drawing characters, and characters from non-US
1091 alphabets fit into this category. Their use presents
1092 interoperability problems that are best avoided.
1094 The UTF-7 encoding of mailbox names also raises questions about what
1095 to do with the hierarchy delimiters in encoded names: do we encode
1096 each hierarchy level and separate them with delimiters, or do we
1097 encode the fully qualified name, delimiters and all? The answer for
1098 IMAP is the former: encode each hierarchy level separately, and
1099 insert delimiters between. This makes it particularly important not
1100 to use as a hierarchy delimiter a character that might cause
1101 confusion with IMAP's modified UTF-7 [UTF-7 and RFC-2060] encoding.
1103 To repeat: a server should use "/", "\", or "." as its hierarchy
1104 delimiter. The use of any other character is likely to cause
1105 problems and is STRONGLY DISCOURAGED.
1107 3.4.11. ALERT Response Codes
1109 The protocol spec is very clear on the matter of what to do with
1110 ALERT response codes, and yet there are many clients that violate it
1111 so it needs to be said anyway: "The human-readable text contains a
1112 special alert that must be presented to the user in a fashion that
1113 calls the user's attention to the message." That should be clear
1114 enough, but I'll repeat it here: Clients must present ALERT text
1115 clearly to the user.
1122 Leiba Informational [Page 20]
1124 RFC 2683 IMAP4 Implementation Recommendations September 1999
1127 3.4.12. Deleting Mailboxes
1129 The protocol does not guarantee that a client may delete a mailbox
1130 that is not empty, though on some servers it is permissible and is,
1131 in fact, much faster than the alternative or deleting all the
1132 messages from the client. If the client chooses to try to take
1133 advantage of this possibility it must be prepared to use the other
1134 method in the even that the more convenient one fails. Further, a
1135 client should not try to delete the mailbox that it has selected, but
1136 should first close that mailbox; some servers do not permit the
1137 deletion of the selected mailbox.
1139 That said, a server should permit the deletion of a non-empty
1140 mailbox; there's little reason to pass this work on to the client.
1141 Moreover, forbidding this prevents the deletion of a mailbox that for
1142 some reason can not be opened or expunged, leading to possible
1143 denial-of-service problems.
1145 Example:
1147 [User tells the client to delete mailbox BANANA, which is
1148 currently selected...]
1149 C: 008 CLOSE
1150 S: 008 OK done
1151 C: 009 DELETE BANANA
1152 S: 009 NO Delete failed; mailbox is not empty.
1153 C: 010 SELECT BANANA
1154 S: * ... untagged SELECT responses
1155 S: 010 OK done
1156 C: 011 STORE 1:* +FLAGS.SILENT \DELETED
1157 S: 011 OK done
1158 C: 012 CLOSE
1159 S: 012 OK done
1160 C: 013 DELETE BANANA
1161 S: 013 OK done
1163 3.5. A Word About Testing
1165 Since the whole point of IMAP is interoperability, and since
1166 interoperability can not be tested in a vacuum, the final
1167 recommendation of this treatise is, "Test against EVERYTHING." Test
1168 your client against every server you can get an account on. Test
1169 your server with every client you can get your hands on. Many
1170 clients make limited test versions available on the Web for the
1171 downloading. Many server owners will give serious client developers
1172 guest accounts for testing. Contact them and ask. NEVER assume that
1173 because your client works with one or two servers, or because your
1174 server does fine with one or two clients, you will interoperate well
1178 Leiba Informational [Page 21]
1180 RFC 2683 IMAP4 Implementation Recommendations September 1999
1183 in general.
1185 In particular, in addition to everything else, be sure to test
1186 against the reference implementations: the PINE client, the
1187 University of Washington server, and the Cyrus server.
1189 See the following URLs on the web for more information here:
1191 IMAP Products and Sources: http://www.imap.org/products.html
1192 IMC MailConnect: http://www.imc.org/imc-mailconnect
1194 4. Security Considerations
1196 This document describes behaviour of clients and servers that use the
1197 IMAP4 protocol, and as such, has the same security considerations as
1198 described in [RFC-2060].
1200 5. References
1202 [RFC-2060] Crispin, M., "Internet Message Access Protocol - Version
1203 4rev1", RFC 2060, December 1996.
1205 [RFC-2119] Bradner, S., "Key words for use in RFCs to Indicate
1206 Requirement Levels", BCP 14, RFC 2119, March 1997.
1208 [RFC-2180] Gahrns, M., "IMAP4 Multi-Accessed Mailbox Practice", RFC
1209 2180, July 1997.
1211 [UTF-8] Yergeau, F., " UTF-8, a transformation format of Unicode
1212 and ISO 10646", RFC 2044, October 1996.
1214 [UTF-7] Goldsmith, D. and M. Davis, "UTF-7, a Mail-Safe
1215 Transformation Format of Unicode", RFC 2152, May 1997.
1217 [NAMESPACE] Gahrns, M. and C. Newman, "IMAP4 Namespace", Work in
1218 Progress.
1220 6. Author's Address
1222 Barry Leiba
1223 IBM T.J. Watson Research Center
1224 30 Saw Mill River Road
1225 Hawthorne, NY 10532
1227 Phone: 1-914-784-7941
1228 EMail: leiba@watson.ibm.com
1234 Leiba Informational [Page 22]
1236 RFC 2683 IMAP4 Implementation Recommendations September 1999
1239 7. Full Copyright Statement
1241 Copyright (C) The Internet Society (1999). All Rights Reserved.
1243 This document and translations of it may be copied and furnished to
1244 others, and derivative works that comment on or otherwise explain it
1245 or assist in its implementation may be prepared, copied, published
1246 and distributed, in whole or in part, without restriction of any
1247 kind, provided that the above copyright notice and this paragraph are
1248 included on all such copies and derivative works. However, this
1249 document itself may not be modified in any way, such as by removing
1250 the copyright notice or references to the Internet Society or other
1251 Internet organizations, except as needed for the purpose of
1252 developing Internet standards in which case the procedures for
1253 copyrights defined in the Internet Standards process must be
1254 followed, or as required to translate it into languages other than
1255 English.
1257 The limited permissions granted above are perpetual and will not be
1258 revoked by the Internet Society or its successors or assigns.
1260 This document and the information contained herein is provided on an
1261 "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
1262 TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
1263 BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
1264 HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
1265 MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
1267 Acknowledgement
1269 Funding for the RFC Editor function is currently provided by the
1270 Internet Society.
1290 Leiba Informational [Page 23]