Network Working Group L. Dusseault Internet-Draft Xythos Expires: May 1, 2004 November 2003 Calendar Server Extensions for WebDAV (CalDAV) draft-dusseault-caldav-00 Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http:// www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on May 1, 2004. Copyright Notice Copyright (C) The Internet Society (2003). All Rights Reserved. Abstract In the five years since [WebDAV] was standardized, at least three groups have used WebDAV as a basis to provide Internet calendar access with a minimum of development effort. This draft explores the reasons why this path might be chosen as well as the mechanisms that might enable interoperable calendar access over WebDAV. Dusseault Expires May 1, 2004 [Page 1] Internet-Draft caldav November 2003 1. Introduction 1.1 Advantages of WebDAV for Calendar Access WebDAV offers a number of advantages as a framework or basis for calendar access. Most of these advantages boil down to a significant reduction in design costs, implementation costs, interoperability test costs, deployment costs, and the costs of mistakes. Every new standard author or implementor finds certain small errors and the IETF spends considerable time and effort remediating these. Some of the advantages are contingent upon the way WebDAV is used, which is why this section exploring advantages is inseparable from the rest of this document for the moment. This draft was commissioned at the Fall 2003 Minneapolis working group meeting of the CalSched working group. The concept of using HTTP/WebDAV as a basis for a calendaring server is by no means a new concept: it was discussed in the CalSched working group as early as 1997 or 1998. Several companies have implemented calendaring servers using HTTP PUT/GET to upload and download iCalendar events, and using WebDAV PROPFIND to get listings of resources. However, those implementations do not interoperate because there are many small and big decisions to be made in how to model calendaring data as WebDAV resources and properties, as well as how to implement required features that aren't already part of WebDAV. This draft is therefore intended as an exploration of the advantages of using WebDAV as well as a proposal for one way to model calendaring data, and some ideas for how to specify the features that go beyond WebDAV. 1.1.1 HTTP URLs for Calendar Objects WebDAV is an extension of [HTTP], therefore its URLs are HTTP URLs. If calendar access were an extension of WebDAV then it could also share HTTP URLs. This can make a lot of sense because it allows very simple calendar browsing clients to be written for devices that already have a HTTP stack: the client merely needs to download those calendar objects and be able to parse their formats. Since the [iCal] formats are well-defined and well-supported, there's a natural choice for what resource to download for a granular calendar object. If HTTP GET can be used to represent a calendar object, then appointment references can be easily downloaded, synchronized and shared. Specifying URLs is work. Not only are HTTP URLs appropriate for Calendar objects, but they also eliminate the need to specify a new URL schema and format and implement it. Dusseault Expires May 1, 2004 [Page 2] Internet-Draft caldav November 2003 1.1.2 Web Services and Web Interfaces Calendar functionality is found extremely frequently on the Web. Even calendaring systems designed primarily for access by smart clients (smart clients are those which have application logic, as opposed to thin clients or Web browsers) typically also have a Web interface accessible by thin clients. Some calendaring applications are available only via Web interfaces, for example those found on systems such as Yahoo! Groups. Because of the frequent use of Web interfaces, and the possibility of supporting Web services, WebDAV is a particularly suitable framework for calendar data. HTTP URLs to calendar objects can be used natively in these systems. WebDAV provides property information in an XML format, easily consumed by Web services which usually import XML data anyway. Web interfaces can use stylesheets to transform XML data into HTML presentation. This approach is described in . 1.1.3 Client Implementations from Simple to Rich The HTTP/WebDAV feature model encourages a wide range of clients, from extremely simple to very rich. This is because servers must support a wide range of features, but clients can pick and choose which features to support. For example, even though a WebDAV server must support the 'lockdiscovery' property, there's no requirement for a client to request or parse this property value if it has no need to. Generally speaking, clients may pick and choose which methods and properties to support, as long as the client has a reasonable response to the error conditions which might be returned. A simple client can merely download and upload iCal objects and use very little XML or advanced WebDAV functionality. At the other end of the scale, a rich calendaring client using WebDAV-based calendaring could choose to implement offline functionality, free-busy searches crossing multiple servers, advanced tasks and even some workflow, by using more of the features and possibly defining its own dead properties. (Note: WebDAV's 'dead' properties are those which the server allows clients to set but the server has no special behavior regarding those properties. Other clients may query and use these dead properties.) 1.1.4 Support for lock feature WebDAV includes locking support. Locks are indispensible when multiple authors may modify or create the same resources. Locks not only prevent authors from accidentally overwriting each others work Dusseault Expires May 1, 2004 [Page 3] Internet-Draft caldav November 2003 (as ETags do), they also help authors coordinate that work by seeing when to wait for another author to finish. Calendar users benefit slightly from this functionality, more so when group calendars or shared calendars allow booking of large groups of people or broadly-used resources such as conference rooms or equipment. 1.1.5 Support for access control The WebDAV ACL specification [ACL] was recently approved by the IESG. However, several implementations have already shown interoperability using this proposal. Any shared or group calendar application benefits from interoperable access control. Access control can help define who can schedule a user for new appointments without having to make email requests, who can view free/busy time, and who can see the details of certain appointments. The IETF Security Area directors are concerned about a flaw in WebDAV ACLs in that different servers can implement different sets of privileges which can be hard for clients to handle. This can be mitigated by a well-written spec for calendaring access via WebDAV, by requiring servers advertising support for Calendaring features to also support a specific and more well-defined set of privileges. That approach is consistent with interoperating with existing WebDAV clients. As a general mechanism, WebDAV ACL may not always provide all the power and specificity required by a calendar application. For example, a calendaring application may wish to choose which other users can view the start/end times of appointments, and separately choose which users can also see the location of appointments. However, as a standard and framework, WebDAV ACL provides a valuable base from which to work. Furthermore, this proposal recommends that advanced access control work for calendaring be relegated to another document, so that standard calendaring systems can be built using existing WebDAV ACL support. 1.1.6 Security, Implentations and Deployed Base Many WebDAV client appliations, servers and APIs already exist. WebDAV clients exist for modern Microsoft, Unix and Apple platforms. Open source solutions are common and powerful. This can significantly improve early interoperability and reduce development and test time. Much security integration work has already been done for WebDAV. Today's Web and WebDAV servers all support TLS, providing at a minimum single-hop privacy and server authentication. HTTP Digest and Basic authentication may provide adequate client authentication Dusseault Expires May 1, 2004 [Page 4] Internet-Draft caldav November 2003 (Basic essentially uses a clear-text password but this may be appropriate if the connection is secured with TLS). If not, work is under way to support SASL with HTTP. As that work nears completion, HTTP/WebDAV implementations will add SASL support so that work will be done already for a calendaring system. It seems the HTTP/SASL work is nearing last call (currently draft-nystrom-http-sasl-09.txt). 1.1.7 Migration, Synchronization and Offline Functionality Synchronization and offline functionality are useful features in Calendaring systems. Luckily, these are already well understood for HTTP/WebDAV technology. HTTP ETags provide a reliable way to determine whether a document in an offline cache needs to be synchronized. At least two WebDAV clients supporting synchronization have already been created: sitecopy (http://www.lyra.org/sitecopy/) and Xythos WebFile Client (http://www.xythos.com/home/xythos/ wfc_features.html). Many WebDAV working group members are discussing more work to improve the performance of synchronization betweeen WebDAV clients and WebDAV repositories. This ongoing work can benefit the calendaring community at the same time, provided that the calendaring data model fits easily in the WebDAV data model. The model proposed in this document is one with which new WebDAV synchronization features are likley to be equally applicable to calendaring data. Data migration is almost the same problem as synchronization. One use of a WebDAV tool like sitecopy is to move data to a new server. The move is peformed by doing a new synchronization. Once the initial synchronization is complete and verified, the data on the old system can be removed or archived. Data portability is a convenient feature to administrators, particularly when deploying a new system. 1.1.8 Search Support Calendaring systems need a mix of fixed, specific searches (such as a "search" for the events occuring today) and general search support. WebDAV DASL can provide the functionality for the general search support (although not always for specific frequently used searches). The only hitch is that DASL is not yet standardized. The WebDAV WG is currently putting effort into completing DASL and several interoperable implementations already exist. In the meantime, if DASL is delayed the specific fixed searches defined in this document (using the REPORT method, see section 7), together with the ability to browse calendars and request calendar objects with certain property values, ought to provide quite reasonable calendar browse/ search support. Dusseault Expires May 1, 2004 [Page 5] Internet-Draft caldav November 2003 Note that the property promotion proposed in this document means that not only can iCalendar documents be searched with "contains" text searches, but also more sophisticated value matching can be done. For example, since 'dtstart' is promoted from a VEVENT document body to the resource's property list, a DASL search can be constructed to find events with 'dtstart' before a specified date. 1.1.9 Clear extensibility model WebDAV has a clear and proven extensibility model. The major way functionality is extended is by defining new properties. Servers can extend functionality by creating new live properties in custom namespaces. Clients can also extend functionality by creating new dead properties in custom namespaces. For example, a client might wish to add a "source-device" property in a custom namespace to record which device created the calendar item. Dead properties are client-controlled properties, where the namespace, name and value are entirely controlled by the client. However, the server is required to store these properties and return them, if requested, in PROPFIND queries for individual resources or in listings of collection contents. Some servers support text searching on all dead properties through the DASL extensions. Dead properties can also be used in reports. Other proven HTTP/WebDAV extensibility mechanisms include the ability to define and advertise special WebDAV reports, new HTTP headers, and for ultimate flexibility, new HTTP methods. 1.2 Required Calendaring Features Missing in WebDAV Certain features that are required for modern enterprise-level calendar systems are not present in HTTP or WebDAV, unsurprisingly. This section makes clear what those missing features are. The rest of this document is largely about how those missing features could be filled in. fanout One of the valuable functions a calendaring server can provide to a client is fanning out scheduling requests. This allows clients to perform simple operations (send a single schedule request to the server) and have the server perform complex and possibly time consuming processing. The server can send individual scheduling requests to different invitees. It can choose the most appropriate scheduling request transport for each invitee, not necessarily using the same technology for each invitee to the same meeting. The server can then remain online to collect and collate responses. Fanout is unlike any existing WebDAV feature. Thus, support for fanout dictates that a Calendaring server cannot Dusseault Expires May 1, 2004 [Page 6] Internet-Draft caldav November 2003 simply be an out-of-the-box WebDAV server with no modifications. A CalDav server must have fanout logic in addition to its basic WebDAV feature support. This document defines how fanout works. recurrance Recurring appointments are common in calendaring applications, and these are the hardest events to fit into a finite resource listing model such as WebDAV. However, WebDAV does have the functionality to perform reports. Recurrance expansion is much like the expand-property report defined in [DeltaV]. It seems pretty straightforward to define a CalDav specific report to do recurrance expansion in the cases where it's most useful. notifications HTTP has no way for the server to contact the client. HTTP interactions all begin with a client request and end with a server response. The server has no obligation to maintain an open connection to the client, rather the server MAY sever the TCP connection after any response. Thus, there is no way for the server to send the client alarms. Since HTTP is not designed for this, it is recommended that a different technology be used to solve this problem. There are other IETF technologies much more suitable than HTTP to provide notifications. Currently, SIP and XMPP are appropriate technologies for servers to send notifications to subscribed clients. The SIP SUBSCRIBE method defined in [REF - Roach RFC] allows custom event types to be declared and subscribed to, and CalDAV (or a separate draft) could define a list of types that CalDAV servers could optionally (or be required to) support. Similarly, XMPP allows clients to Using a separate technology for notifications has advantages. It can be seen as an advantage to have notification support optional, at least for client implementations - a simple CalDav client need not support notifications at all if it doesn't use alarms, or generates its own alarms, or stays offline most of the time. Another advantage is that it allows CalDAV servers to be stateless, handling HTTP requests by generating responses, plus events that can be fired off to a separate notification engine (perhaps a separate server) and then forgotten. Finally, a separate notification system can then deal with scaling to handle the combinatorial problems of multiple subscribed clients (e.g. my PDA, my laptop and my desktop) combined with multiple event sources (e.g. my work calendar server plus my sports team calendar server plus non-calendaring event sources). Dusseault Expires May 1, 2004 [Page 7] Internet-Draft caldav November 2003 It is likely that CalDAV servers will be required to support notification once we have more input on which technology to use. At the least, the server would support notification of new resources added to collections, so that clients can subscribe to know when new iTIP requests have been delivered or when other clients have added new events or other calendaring objects. A CalDAV server might also initiate notification when a reminder is needed. This could allow simple non-calendaring clients (e.g. a cell phone) to subscribe only to reminder event notifications without having to do any other part of calendaring. Dusseault Expires May 1, 2004 [Page 8] Internet-Draft caldav November 2003 2. Required CalDAV features This section lists what functionality is required of a CalDAV server. To advertise support for CalDAV or claim compliance, a server: o MUST support WebDAV Level 1 and 2 (all of RFC2518 including locking). o MUST support WebDAV ACL. See section (TODO) for more specifics on how o MAY support WebDAV DeltaV [DeltaV]. o MUST support DASL. o MUST support HTTP/SASL. o MUST support property promotion as described in this document. o MUST support scheduling fanout as described in this document. o MUST support calendaring REPORTs as described in this document. Dusseault Expires May 1, 2004 [Page 9] Internet-Draft caldav November 2003 3. Calendaring Data Model One of the features which has made WebDAV a successful protocol is its firm data model. This makes it a useful framework for other applications such as calendaring. In this proposal, we attempt to follow the same pattern by developing all new features based on a well-described data model. In this data model, every iCalendar VEVENT, VALARM, VJOURNAL, VTODO and VFREEBUSY is stored as a regular HTTP/WebDAV resource. That means each calendar resource may be individually locked and have individual properties. These resources are sorted into WebDAV collections with a mostly-fixed structure. All resource types defined in this section, and all resources of MIME type text/calendar appearing within these collections, have certain required properties. These properties may be defined in this document or in a separate document [TODO. Note: OSAF uses a calendaring property model based on RDF calendaring and iCal that could fit here]. Another TODO: This section needs more information on what properties are REQUIRED on each type of collection. The iTIP document has useful tables listing properties for each method, which might apply to these collections. 3.1 Calendar Repository or Server A CalDav repository, or server, is a calendaring-aware engine combined with a WebDAV repository. A WebDAV repository is a set of WebDAV collections, containing other WebDAV resources, within a unified URL namespace. For example, the repository "http:// example.org/webdav/" may contain WebDAV collections and resources, all of which have URLs beginning with "http://example.org/webdav/". Note that the root URL "http://example.org/" may not itself be a WebDAV repository (for example, if the WebDAV support is implemented through a servlet or other Web server extension). A WebDAV repository may include calendar data in some areas, and non-calendaring data in other areas. Calendar data will be indicated through specific container relationships and resource types discussed in the next sections. A WebDAV repository may advertise itself as a CalDAV server if it supports the functionality defined in this specification at any point within the root of the repository. That might mean that calendaring data is spread throughout the repository (e.g. in /ldusseault/ calendar as well as in /nborenstein/calendar) or only in certain sections of the repository (e.g. /caldav/usercals/*). Calendaring features are only required in the repository sections that are or Dusseault Expires May 1, 2004 [Page 10] Internet-Draft caldav November 2003 contain calendaring objects. So a repository confining calendar data to the /caldav/ collection would only need to support calendaring REPORTs defined here within that collection. The CalDAV server or repository is the canonical location for calendar data, state information and semantics. The CalDAV server has significant responsibility to ensure that the data is consistent and compliant. Clients may submit requests to change data or download data. Clients may store the calendar offline and attempt to synchronize when reconnected, but changes to the repository occurring in between are not considered to be automatically disposable and clients should consider the repository to be the first authority on state. HTTP Etags and other tools help this work. 3.2 Calendar Containers A WebDAV collection which contains one or more calendars is considered a Calendar Container. It has a new resource type: A calendar container may contain more than only Calendar resources. However, non-Calendar resources within a Calendar-Container are not typically intended for user display. These resources may contain configuration or application data created by clients or offered by the server for use by clients. 3.3 Calendars A WebDAV collection which corresponds to a single calendar or VAGENDA is a Calendar. It has a new resource type: The calendar collection contains sub-collections with specific kinds of calendar objects. It also has certain properties which are required to be present on calendards (see XML section). Calendars MUST NOT contain other calendars. Calendars MAY exist inside calendar-containers or inside normal WebDAV collections. Thus, a repository may have calendars without having calendar-containers. Calendar-containers are typically useful so Dusseault Expires May 1, 2004 [Page 11] Internet-Draft caldav November 2003 that a client can automatically detect when a user has multiple calendars, e.g. "/ldusseault/calendars/work" and "/ldusseault/ calendars/karate". A Calendar has a specified substructure. It MUST contain one event collection and one alarm collection. It MAY contain one todo collection and one journal. It MUST NOT contain more than one of any of these specific collections, although it MAY contain additional collections and non-collection resources of types not defined here. 3.4 Event Collection Each Calendar MUST have a collection containing events. All resources within this event collection (even within its sub-collections) are considered part of the calendar, so substructure can be used to organize events into smaller collections without affecting the overall content of the calendar. The event collection has its own resource type. Every non-collection resource in a calendar-events collection is considered to be an event. Thus, listing the resources inside a calendar-events collection is a good way to find out all the events on a calendar. Each resource inside an events collection MUST have the default MIME type text/calendar, and each one contains exactly one VEVENT or VFREEBUSY object. 3.5 Alarm Collection Each Calendar MUST have a collection containing alarms. All resources within this alarm collection (even within its sub-collections) are considered part of the calendar. The alarm collection has its own resource type. Every non-collection resource in a alarms collection is considered to be an alarm. Every resource MUST have the default MIME type text/ calendar, and contains exactly one VALARM. Dusseault Expires May 1, 2004 [Page 12] Internet-Draft caldav November 2003 3.6 Todo Collection Each Calendar MAY have a collection containing tasks or todos. All resources within this todo collection (even within its sub-collections) are considered part of the calendar. The todo collection has its own resource type. Every non-collection resource in a todo collection is considered to be a todo. Every resource MUST have the default MIME type text/ calendar, and contains exactly one VTODO. 3.7 Journal Each Calendar MAY have a collection containing journal items. All resources within this journal collection (even within its sub-collections) are considered part of the journal. The journal collection has its own resource type. Every non-collection resource in a journal is considered to be a journal item. Every resource MUST have the default MIME type text/ calendar, and contains exactly one VJOURNAL. 3.8 Invites Collection Every Calendar-Container MUST have a child collection to contain invites received through iTIP. If a Calendar is not inside a Calendar-Container, then that Calendar MUST have its own invites collection. This collection is to store received, accepted and archived invitations so that clients can accept and review scheduling invitations. Every non-collection resource in the invites collection is considered to be a REQUEST. Every resource MUST have the default MIME type Dusseault Expires May 1, 2004 [Page 13] Internet-Draft caldav November 2003 text/calendar, and contains exactly one REQUEST. 3.9 Scheduling Collection Every Calendar MUST have a child collection to contain fanout requests and responses for appointments scheduled by the calendar owner (or other users of this calendar). This collection is to store REQUESTs initiated by this calendar server for this calendar, as well as REPLY items received in reply. This collection is only for review because the CalDAV server is responsible for parsing incoming REPLY messages and adding attendee information to events. Every non-collection resource in the invites collection is considered to be a REQUEST or REPLY. Every resource MUST have the default MIME type text/calendar, and contains exactly one REQUEST or exactly one REPLY. 3.10 Recurrance and the Data Model Recurrance is an important part of the data model because it governs how many resources are expected to exist. Consider the outcome if recurrance were handled through the creation of many nearly-identical WebDAV resources. With this model, it becomes hard to keep their data consistent. Even worse, some features like LOCK become difficult -- it's hard to lock the right set of resources so that the user can change the title of all recurrances of an appointment. With these considerations, this proposal does not treat recurrances as separate resources. Instead, this proposal models recurrance patterns as properties of event resources. This makes for much less data to synchronize, and makes it easier to make changes to all recurrances or to a recurrance pattern. It makes it easier to create a recurring event, and easier to delete all recurrances. The drawback of the recurrance-is-a-property approach is that it becomes harder to see what events occur in a given time period. It's a very common function for calendar views to display all events happening between midnight yesterday and midnight tonight, or all events happening within one week. In these views, each recurrance appears as if it were an individual appointment. To make these views possible, this proposal defines a REPORT specifically to view events Dusseault Expires May 1, 2004 [Page 14] Internet-Draft caldav November 2003 in a time period [TODO - ref section]. Because of this choice, clients MUST NOT create separate resources to represent a recurring event when the recurrance pattern is known. Otherwise, it makes it more difficult for other clients to interoperate and modify the recurring event. Most importantly, clients MUST NOT duplicate events represented through recurrance patterns with manually created events, which would appear as duplicates to the server and to other clients. 3.11 Calendar, event and custom timezones This is a big issue because CalDAV servers need to know the timezone of the calendar in order to calculate recurrances in a deterministic manner. In addition to having a calendar default timezone, individual events (e.g. those received in a scheduling request from another place) can have different timezones. Finally, iCalendar has a way to specify custom timezones. The default calendar timezone is probably simply a property value on the calendar collection, which the calendaring client can change. The event timezone is a piece of VEVENT metadata that would therefore appear both inside the event in iCalendar format, and as a promoted property on the event resource. Storing custom timezone definitions can be through a new collection if necessary. More discussion here is welcome. 3.12 Creating Resources Calendars, calendar-containers, collections of calendar objects, and individual calendar objects may all be created by either the CalDAV client or by the CalDAV server. For example, a server might come preconfigured with a user's calendar collection, or the CalDAV client might create a new calendar collection. Servers might create event requests as calendar objects inside a VEVENT collection, or clients might create event requests. Either way, both client and server MUST comply with the requirements in this document, and MUST understand objects appearing in calendars or calendar-containers according to the data model defined here. When servers create HTTP resources, it's not hard for the server to choose a unique URL. It's slightly tougher for clients, because a client might not want to examine all resources in the collection, and might not want to lock the entire collection to ensure that a new one isn't created with a name collision. However, there are tools to mitigate this. If the client intends to create a new non-collection resource, such as a new VEVENT, the client SHOULD use the HTTP header "If-None-Match: *" on the PUT request. The Request-URI on the PUT Dusseault Expires May 1, 2004 [Page 15] Internet-Draft caldav November 2003 request MUST include the target collection, where the resource is to be created, plus the name of the resource in the last path segment. The last path segment could be a random number, or it could be a sequence number, or a string related to the object's 'summary' property. No matter how the name is chosen, the "If-None-Match" header ensures that the client cannot overwrite an existing resource even if it has accidentally chosen a duplicate resource name. PUT /lisa/calendar/events/mtg10028.ics HTTP/1.1 If-None-Match: * Host: cal.example.com Content-Type: text/icalendar Content-Length: XXX BEGIN:VEVENT DTSTART:20010714T170000Z DTEND:20010715T035959Z SUMMARY:Bastille Day Party END:VEVENT The request to change an existing event is the same, but with a specific ETag in the "If-Match" header, rather than the "If-None-Match" header. For optimum interoperability with existing HTTP clients, CalDAV clients and servers MUST use the file extension ".ics" as well as the "text/ icalendar" MIME type, whenever creating Calendar objects of that MIME type. Note because of these requirements that there is no semantic value in any other part of a resource name other the file extension. Thus, a Calendar collection may be called "calendar", "cal", "Calendario" or "日历" (Chinese). It's the properties of the resource that define what it is, not the name. 3.13 Users and Groups The WebDAV ACL specification requires that any principal to whom permissions can be represented via a WebDAV resource (complete with WebDAV properties and a HTTP URL). Thus, both users may be represented (for example, as /principals/users/ldusseault) and groups (for example, as /principals/groups/calsch). This feature offers an excellent framework for linking users to calendars in a fashion not otherwise easily implemented. Note that the WebDAV principal resources may not be modifiable through WebDAV. This is an important consideration because it allows the principal directory to be merely a WebDAV representation of data Dusseault Expires May 1, 2004 [Page 16] Internet-Draft caldav November 2003 which is canonically stored in an outside system. For example, an enterprise might use an LDAP server to store and administer all user and group properties. This LDAP server could be linked into the WebDAV repository through configuration information. WebDAV server implementations exist which offer principal resources, but when the principal resources are queried the server actually makes a LDAP request to get the principal information from its official source. This saves WebDAV clients from having to implement LDAP and provides a single URL format for principals regardless of whether the user directory is stored in LDAP or some other system. A server supporting CalDAV MUST support additional properties on principal resources if these principals are associated with calendars. In addition, certain properties are required on calendars to link to principal resources. These properties are defined in the properties section. Dusseault Expires May 1, 2004 [Page 17] Internet-Draft caldav November 2003 4. Property Promotion and Demotion Property promotion and demotion (hereafter called simply "property promotion" is the name for the functionality by which a server ensures that a resource's internal data and its externally-visible metadata remain consistent. In WebDAV, a collection listing (PROPFIND) selects a set of property names to retrieve. For a collection listing to be useful to browse calendars, certain calendaring information must be exposed as WebDAV properties (this also makes WebDAV SEARCH useful, and makes the definition of REPORTs easier). Since a calendar resource of type text/calendar has properties which duplicate some of its internal state, it's the server's responsibility to keep those consistent somehow. The server has some leeway in how it makes properties and bodies consistent, as long as the response to a GET shows information consistent with the response to a PROPFIND in the interval in which a calendar object has not been altered. Thus, the server MAY change property values when a PUT is performed that alters data exposed as properties, and also change the body when a PROPPATCH is performed that alters calendar properties. Alternatively, a server could implement "lazy promotion" and apply consistency changes only when a GET, PROPFIND, SEARCH or REPORT is issued. Finally, a server might decompose property data and non-property data into separate locations and recompose the information only when a GET requests the entire resource. Any of these approaches MUST be transparent to the client, in that operations behave consistently, with complete round-trip fidelity of all the data originally provided. Thus, a server MAY canonicalize its resource bodies (e.g. eliminate meaningless spaces) but MUST preserve all data. This section needs further definition and details. Clients can upload iCalendar files with syntactic or semantic errors, so helpful error codes must be chosen. Also, this draft needs to define what to do with 'x-' custom elements inside these uploaded files. Dusseault Expires May 1, 2004 [Page 18] Internet-Draft caldav November 2003 5. Scheduling and Fanout Scheduling and fanout is a valuable function provided by advanced calendaring servers. Simple clients clearly benefit from having the logic handled by the server. Rich clients also benefit from having to upload less data to various servers (including messaging servers to send invitations via messages) to accomplish the same things. Servers can sometimes provide more advanced scheduling functionality than clients - for example, a server providing fanout could create "unconfirmed" VEVENT resources within invitees' calendars. However, rich calendaring clients may prefer to do fanout. Clients can perform special functionality during scheduling (for example, a client may be configured to be able to directly put events on others' calendars if the user has sufficient permissions). Thus, it is proposed that CalDAV allow the client to either perform fanout and merely create the event (complete with attendee information) OR request that the server perform fanout. In other words, the server MUST handle fanout if requested, and clients MAY perform fanout if the client chooses. CalDAV servers MUST support iTIP to do fanout when requested, and MUST fall back to iMIP when iTIP is impossible. Each REQUEST generated on behalf of a Calendar MUST be stored in the calendar's 'scheduling' collection. Each REPLY received to those REQUESTs MUST be stored in the same collection. These REQUEST and REPLY items are available until the client deletes them so that clients can view extra information (if present). CalDAV servers MUST parse incoming REPLY messages and update the appropriate event with attendee information. Thus, it's not necessary for clients to review REQUEST or REPLY items, although they may. As well, CalDAV servers MUST support iTIP to receive scheduling requests. When a CalDAV server receives an iTIP request, it MUST store the REQUEST object in an "invites" collection for the client to handle. Each invite resource will have properties indicating whether it is new, has been accepted, has been rejected, and whether it is an obsolete REQUEST (the event has passed). Note that when a calendar server receives iTIP requests it MAY auto-accept based on user configured preferences. How these preferences are configured is out of the scope of this specification, but one could imagine that a CalDAV server could host auto-accept configuration Web pages. A CalDAV server is NOT REQUIRED to do any auto-accepting, it MAY simply store the requests for the next time the client is online. Exact mechanisms for triggering fanout requests must be determined and input is welcome. There are several ways fanout could be accomplished: (a) A PUT of the resource triggers fanout, so the body Dusseault Expires May 1, 2004 [Page 19] Internet-Draft caldav November 2003 must contain the fanout information (text and flags), (b) a PROPPATCH triggers fanout if certain properties are set, (c) a new method requests fanout of a resource that has already been uploaded. These three approaches are the most obvious to this author and there is surprisingly little to choose between. More input is needed, for example input on whether the fanout should be synchronous or asynchronous. An asynchronous fanout mechanism using PUT or PROPPATCH would mean that the client would synchronously handle the PUT or PROPPATCH itself, but send invitations at some later time. A synchronous fanout mechanism would probably use a new method with a name like INVITE, because adding new synchronous behavior to existing methods might require more complicated server implementation work. When the server does fanout, it may send requests and receive replies. Probably these requests and responses should be stored as WebDAV resources so that the client can examine the details if desired. This could be a separate collection within the calendar collection. Dusseault Expires May 1, 2004 [Page 20] Internet-Draft caldav November 2003 6. CalDAV Resource Properties The namespace "urn:ietf:params:xml:ns:calsch" is reserved for specifications written by the CalSched working group or its eventual successors. It MUST NOT be used for custom extensions. It is the namespace for every new defined in this section (and every XML element defined in this document). The W3C RDF Calendar group has already defined a namespace and XML element names for many calendaring properties, and these are completely consistent with iCal [RDF-ICAL]. This standard reuses those namespaces, names and definitions. Additional properties are needed to describe calendars and calendar-containers, because the W3C RDF Calendar group defines properties for the iCalendar-defined objects only. The properties that are defined by the RDF Calendar group are not redefined in this section. Note that the XML Schema declarations used in this document are incomplete, in that they do not include namespace information. Thus, the reader MUST NOT use these declarations as the only way to create valid CalDAV properties or to validate CalDAV-related XML. Some of the declarations refer to XML elements defined by WebDAV which use the "DAV:" namespace. Those WebDAV elements are not redefined in this document. 6.1 Required Properties: VEVENT TODO: Has the RDF Calendar group, or iCal, already done the work of defining which properties MUST exist on a VEVENT object? This should just be reused. Required properties probably include: In the 'http://www.w3.org/2002/12/cal/ical' namespace: dtstart dtend or duration... TODO - perhaps a separate document on 6.2 Calendar-owner Property Name: calendar-owner Location: MUST appear on a calendar or calendar-container if there is a principal resources (user or group) with which it is associated. Purpose: This property is used to discover the calendar owner, or the person or group for which the calendar events are scheduled. It MAY have the same value as the DAV (ACL) 'owner' property, but not Dusseault Expires May 1, 2004 [Page 21] Internet-Draft caldav November 2003 necessarily. For example, the president may be the calendar-owner, but her secretary is the owner for all ACL/ administrative purposes. This property contains one 'href' element in the "DAV:" namespace. Declaration: Extensibility: MAY contain additional elements, which MUST be ignored if not understood. 6.3 Calendars Property Name: calendars Location: MUST appear on ACL Principal resources with which a calendar is associated. Purpose: This property is used to browse principals and find out what calendars they have, for the purpose of then browsing their calendar or their free/busy published information. It contains one or more 'href' elements in the "DAV:" namespace (this is a normal WebDAV approach to putting one or more URL value in a property value). Declaration: Extensibility: MAY contain additional elements, which MUST be ignored if not understood. Dusseault Expires May 1, 2004 [Page 22] Internet-Draft caldav November 2003 7. Calendaring Reports This section defines the reports which a CalDAV server MUST support. These all provide special query functionality not normally handled by the generic PROPFIND or SEARCH mechanisms. This can be required when a PROPFIND or SEARCH cannot be written to request the data required for a common use case without an reasonable amount of complex calculation or unnecessary data transmitted. See DeltaV or ACL standards for some examples of reports required in other situations. As defined in DeltaV, all REPORT requests include an XML body naming the type of report requested (only one) and some variables for how that report is to be compiled. Note that support for the REPORT method does not imply support for all reports defined in all WebDAV extensions. A CalDAV server is required to support all the reports defined here and in the ACL standard, but is not expected to support DeltaV reports unless it advertises them. Reports are advertised with the 'supported-report-set' property defined in DeltaV (again, even if versioning-specific features are not at all supported). Each report defined here comes with specialized errors. In addition, some WebDAV status codes are applicable to any request or to any REPORT request. This includes redirect status codes, syntax errors (400 Bad Request), permission errors or policy errors (401 Unauthorized and 403 Forbidden), 404 Not Found, or a request-body that isn't XML or is invalid XML (422 Unprocessable Entity). When an error is defined in this document, it is used in an error response body inside an XML document (this practice was established with DeltaV and ACL in order to avoid status code collisions). For example: Sample error response HTTP/1.1 409 Conflict Date: Sun, 16 November 2003 18:40:01 GMT Content-Type: text/xml; charset="utf-8" Content-Length: XXX 7.1 The 'calendar-time-range' report The 'calendar-time-range' report returns all objects of a specific Dusseault Expires May 1, 2004 [Page 23] Internet-Draft caldav November 2003 type within a time range, with or without recurrance expanded. The first use case for this report is to have the server expand recurring events to make a calendar view of a day's or week's events easy. The WebDAV PROPFIND and SEARCH syntaxes do not as easily support this use case. Even when the client doesn't need recurrance expanded, it can use this report to save itself from the need to write a SEARCH query which catches all events overlapping any part of the period requested, or from having to do a PROPFIND and filter itself. The second use case for this report is for users other than the calendar owner to find out when the calendar owner is free. This is only a minor variation, because it's effectively the same objects (VEVENT and VFREEBUSY), only with permissions restricting the kind of data the server will return. Servers MUST allow users with permission to view the free-busy times for a calendar to use this report. Servers MUST return event properties for visible events including dtstart, dtend and free-busy type. Other properties MAY be refused. The third use case for this report is to list all alarms in a time range. The selection of VALARM objects, instead of VEVENT or VFREEBUSY objects, allows this use case to be handled with the same report framework. 7.1.1 Request for 'calendar-time-range' The REPORT request-body MUST have the root element 'calendar-time-range'. o The root element MAY contain the 'expand-recurrances' element as a flag. o The root element MAY contain the 'object-types' element to list what object types to return. o The root element MUST contain the 'prop' element in the "DAV:" namespace as defined in WebDAV, to list what property values to return. o The root element MUST contain one 'dtstart' element o The root element MUST contain one 'dtend' element. The Request-URI for this report MUST be a Calendar-container, a calendar collection, or an events collection. The server MUST collation all the event data contained within the requested collection. Dusseault Expires May 1, 2004 [Page 24] Internet-Draft caldav November 2003 Sample request for 'time-range-events' report REPORT /lisa/Calendar HTTP/1.1 Host: cal.example.com Content-Type: text/xml Content-Length: XXX 20031101 20031131 7.1.2 Response to 'time-range-events' The response to this report is a WebDAV Multi-Status response, containing one element for each event AND for each recurrance. This differs from the PROPFIND response to an event collection only in that the relevant recurrances each have their own element, not just the master event. The server MUST expand all recurring events within the entire collection (including sub-collections) if requested, and return all those events or recurrances which overlap the period defined by the start to end. If an event ends at precisely the requested start time, or begins at precisely the requested end time, it does not overlap the period requested. If the user requests properties which may not be seen (e.g. a user with permission only to see free-busy time requests to see the location of events), the response uses the regular WebDAV approach for properties which are private (either 401 Unauthorized if the client is not authenticated, or 403 Forbidden if the client is authenticated and still the property value is private). These errors appear within the standard Multi-Status response. TODO: I guess an example is probably needed here. 7.1.3 Errors for 'time-range-events' Dusseault Expires May 1, 2004 [Page 25] Internet-Draft caldav November 2003 invalid-range The server returns this error when the range requested in the 'dtstart' and 'dtend' values is an invalid range (e.g. dtend is earlier than or equal to dtstart value). 7.1.4 Timezones report CAP offers a way to get timezones. I'm unclear whether this is a server-wide list, or whether the contents vary by calendar/user. That would help to decide whether to use a special OPTIONS request, a new REPORT, or a new collection. Dusseault Expires May 1, 2004 [Page 26] Internet-Draft caldav November 2003 8. Using existing WebDAV features in Calendaring 8.1 SEARCH and calendar data The DASL framework for search requests provides a powerful way to find calendars in a repository, and to find calendar objects within a calendar. It is virtually unlimited in variations. It can be used to request and search on calendar properties as well as WebDAV properties. One drawback of DASL, however, is that implementations are given great leeway in which properties support search. That's less acceptable in calendaring applications, so this specification adds requirements of CalDAV servers to support searches on specific properties. CalDAV servers MUST support 'eq' DASL searches on the following properties: uid, valarm. CalDAV servers MUST support 'eq', 'gt' and 'lt' DASL searches on the following properties: dtstart, dtend, dtstamp. CalDAV servers MUST support 'eq' and 'contains' DASL searches on the following properties: location, comment, description, summary, organizer, attendee, categories. 8.2 Disconnected Operations WebDAV already provides functionality required to synchronize a collection or set of collections, make changes offline, and a simple way to resolve conflicts when reconnected. Strong ETags are the key to making this work, but these are not required of all WebDAV servers. Since offline functionality is more important to Calendar applications than to other WebDAV applications, CalDAV servers MUST support strong ETags. Much more work could be done to make disconnected operations work better. WebDAV implementors have discussed ETag-like tags for collections (CTags?) which would change whenever the membership (or members?) of a collection changed. Tombstones might also be useful to synchronize with DELETE operations. However, all these mechanisms are of general use and not limited to Calendaring. Therefore, it is suggested that work on advanced synchronization take place in a separate document independent of the calendaring-specific features discussed here. Many people are interested in doing this kind of work and it has wide applicability and usefulness. Requirements or design contributions from calendaring implementors are welcome. TODO: this section should be expanded to give more guidance to clients on how to synchronize WebDAV objects most effectively. In Dusseault Expires May 1, 2004 [Page 27] Internet-Draft caldav November 2003 particular, we need to understand how UID/SEQ metadata works with synchronization. Note that recurrance isn't a synchronization problem in this model. Recurring items appear only once in normal PROPFIND responses, so there's no danger that in synchronizing a client will accidentally create extra recurrances. Instead, recurrances appear only in a special REPORT which MUST not be used for synchronization. We believe this separation between data (recurring appointments) and presentation (the display of a period containing several recurrances) is crucial to simplifying synchronization. Dusseault Expires May 1, 2004 [Page 28] Internet-Draft caldav November 2003 References [1] Goland, Y., Whitehead, E., Faizi, A., Carter, S. and D. Jensen, "HTTP Extensions for Distributed Authoring - WebDAV", RFC 2518. [2] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., Leach, P. and T. Berners-Lee, "Hypertext Transfer Protocol -- HTTP/1.1", June 1999. [3] Dawson, F. and D. Stenerson, "Internet Calendaring and Scheduling Core Object Specification (iCalendar)", RFC 2445. [4] "iCal Schema in RDF/XML", site http://www.w3.org/2002/12/cal/. [5] Mogul, J., Krishnamurthy, B., Douglis, F., Feldmann, A., Goland, Y., van Hoff, A. and D. Hellerstein, "Delta encoding in HTTP", RFC 3229, January 2002. Author's Address Lisa Dusseault Xythos Software, Inc. 2064 Edgewood Dr. Palo Alto, CA 94303 US EMail: lisa@xythos.com Dusseault Expires May 1, 2004 [Page 29] Internet-Draft caldav November 2003 Appendix A. Acknowledgements Cyrus Daboo and Michael Arick have provided substantial feedback for this draft. Dusseault Expires May 1, 2004 [Page 30] Internet-Draft caldav November 2003 Appendix B. Changes Dusseault Expires May 1, 2004 [Page 31] Internet-Draft caldav November 2003 Intellectual Property Statement The IETF takes no position regarding the validity or scope of any intellectual property or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; neither does it represent that it has made any effort to identify any such rights. Information on the IETF's procedures with respect to rights in standards-track and standards-related documentation can be found in BCP-11. Copies of claims of rights made available for publication and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementors or users of this specification can be obtained from the IETF Secretariat. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights which may cover technology that may be required to practice this standard. Please address the information to the IETF Executive Director. Full Copyright Statement Copyright (C) The Internet Society (2003). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assignees. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION Dusseault Expires May 1, 2004 [Page 32] Internet-Draft caldav November 2003 HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Acknowledgment Funding for the RFC Editor function is currently provided by the Internet Society. Dusseault Expires May 1, 2004 [Page 33]