|WebDAV Working Group||J. Stracke, Jim Amsden|
|INTERNET DRAFT||Netscape, IBM|
|<version-goals-01221999.htm>||January 22, 1999|
|Expires July, 1999|
This document is an Internet-Draft. 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 made obsolete 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."To view the list Internet-Draft Shadow Directories, see http://www.ietf.org/shadow.html.
Distribution of this document is unlimited. Please send comments to the Distributed Authoring and Versioning (WebDAV) working group at email@example.com, which may be joined by sending a message with subject "subscribe" to firstname.lastname@example.org. The main WebDAV mailing list at email@example.com may also be used for comments. It may be joined by sending a message with subject "subscribe" to firstname.lastname@example.org.
Discussions of the WebDAV versioning working group are archived at http://lists.w3.org/Archives/Public/ietf-dav-versioning/. Discussions of the WebDAV working group are archived at http://www.w3.org/pub/WWW/Archives/Public/w3c-dist-auth.
Versioning and configuration managerment are important features for controlling the evolution of remotely authored Web content. Parallel development leverages versioning capability to allow multiple authors to simultaneously author Web content. This document supersedes the versioning-related goals of [WEBDAV-GOALS].
|Status of this Memo|
Versioning, parallel development, and configuration management are important features for remote authoring of Web content. Version management is concerned with tracking and accessing the history of important states of a single Web resource, such as a standalone Web page. Parallel development provides additional resource availability in multi-user, distributed environments and lets authors make changes on the same resource at the same time, and merge those changes at some later date. Configuration management addresses the problems of tracking and accessing multiple interrelated resources over time as sets of resources, not simply individual resources. Traditionally, artifacts of software development, including code, design, test cases, requirements, help files, and more have been a focus of configuration management. Web sites, comprised of multiple inter-linked resources (HTML, graphics, sound, CGI, and others), are another class of complex information artifacts that benefit from the application of configuration management.
The WebDAV working group originally focused exclusively on defining version management capabilities for remote authoring applications, and group consensus on these features is reflected in [WEBDAV-GOALS]. However, as the WebDAV working group has constructed protocols for versioning functionality, it has become clear that while versioning functionality alone is useful for a range of content authoring scenarios involving one, or a small set of resources, versioning alone is insufficient for managing larger sets of content. Protocol support for parallel development and simple remote configuration management of Web resources provides functionality for managing larger sets of interrelated content developed by multiple users at different locations.
This document describes functional goals for versioning, parallel development, and configuration management of Web resources which replace the existing functional goals for versioning capability described in [WEBDAV-GOALS], section 5.9.
This section provides an example usage scenario that provides a context for explaining the definitions above, and for exploring and validating the goals given in the rest of this document. The example consists of a fictious company, Acme Web Solutions, that is developing a typical Web e-business application. To provide for the broadest coverage, the scenarios start with a non-existant resource typical of web applications, and follow its life-cycle through development and multiple deployments. Other resources would likely have similar life-cycles.Acme Web Solutions (AWS) has developed a web-grocery store called WGS. The application consists of a number of HTML pages, some Java applets, some Java Server Pages (JSP) and a number of Java servlets that access a DB2 database.
AWS has decided to develop a new generation of its flagship WGS product to include maintenance of customer profile information, and active (push) marketing of product specials to interested customers using Channel Definition Format (CDF). The new product will be called Active Grocery Store or AGS. Customers who are interested in receiving information on specials will indicate that interest by subscribing to various CDF channels targeting pre-defined, or user-specified product groupings. Since AGS represents significant new revenue potential for grocery stores, AWS has decided to sell it as a separate product from WGS, and at a relatively high price. WGS will still be available without AGS as a lower-cost, entry level solution for smaller stores, or stores just getting into e-business solutions.
AGS is a typical Web application development project that will require changes to existing resources in AWS as well as adding new resources. These new resources will also be HTML pages, applets, JSPs, servlets, etc. WGS is an active project sold to current customers with a maintenance contract. It has on-going updates that are unrelated to the new AGS system, but may need to be included in the AGS system. These include bug fixes or minor new functional improvements. Since AGS is based on WGS, but both can evolve and be sold separately, it is necessary to maintain versions of resources used by both. This will require AWS developers to specify a configuration of versioned resources corresponding to each product. As the products evolve over time, these configurations will be versioned resources themselves, each representing a new release of their associated product, WGS or AGS.
The AWS development organization consists of a large number of developers across a variety of disciplines including webmasters, Java developers, relational database developers, HTML page editors, graphics artists, etc. All of these developers contribute to the development of the WGS and AGS products, often working in parallel on the same resource for different purposes. For example, a WGS developer may be editing an HTML page to fix a usability problem while an AGS developer is working on the same page to add the new AGS functions. This will require coordination of their activities to provide maximum availability of these shared resources while at the same time ensuring the integrity of the updates. AWS has decided to allow parallel development and resolve multiple concurrent updates through branching and merging of the resource version graph. This adds complexity to the development project as well as some risk due to inaccurate merges, but AWS has decided it cannot be competitive in the Web world if all development must be serialized on shared resources as this would significantly slow product development.
The following scenarios trace the life-cycle of a typical Web resource from conception to product deployment and maintenance. Each scenario exposes some aspect of WebDAV and its use of the versioning, parallel development, and configuration management definitions and goals specified in this document. In the scenarios below, it is assumed that all developers have access to a Web WorkBench (WB) application that provides client access to a WebDAV server called DAVServer. It is further assumed that both the client and server provide level 2 WebDAV services plus advanced collections, versioning, parallel development, and configuration management.
The AGS project team held a design meeting to determine the work products required to support the AGS project, its integration with the WGS application, and to assign these work products to developers. Various analysis and design techniques can be used to discover the required work products, but this is beyond the scope of WebDAV. At the end of the meeting, webmaster Joe was assigned to develop the new welcome page, index.html, for the AGS project. This page will be the initial page used to navigate the AGS application, and is the first page seen by users. It is a new page that will not replace the WGS welcome page, but will contain a reference to it.
Joe uses WB to create create a new collection, http://aws/ags/, and the new index.html page in the collection http://aws/ags/index.html. Nether the parent collection, nor index.html are versioned resources at this point. A WebDAV MKCOL is used to create the collection, and a PUT is used to create the initial, empty resource.
Joe uses WB to GET the resource and edit it with his favorite HTML editor. Each save by the HTML editor does a PUT to the DAVServer, overwriting its current contents. No new versions are created. Joe may also use WB to get and set properties of index.html using PROPFIND and PROPPATCH. Joe does not need to lock index.html because he is the only developer working on it at this time. He could however lock the resource to ensure no one else could make any changes he is not aware of.
At some point, Joe decides preliminary editing on index.html is complete, and he needs to make a stable version available to other developers who need it for integration testing, etc. Joe however wants to ensure that no other developers make changes to to index.html that he cannot back out as he is the webmaster responsible for the resource. So Joe uses the WB to check in index.html which causes DAVServer to create a versioned resource, and make the initial or root version Joe's index.html. At this point, Joe's index.html is immutable, it cannot be changed by anyone, including Joe, and remains in the repository until the versioned resource is deleted.
When DAVServer created the versioned resource corresponding to index.html, it gave the initial version a revision id, "102847565". This revision name is automatically assigned by the server, and cannot be changed or assigned to any other version. This revision name acts as the unique identifier for this version of versioned resource index.html. The AGS development team has decided that the initial version of all resources will be identified by a revision label "initial" so that they stand out and can be easily accessed without remembering some opaque revision id. Joe uses WB to set the label on the initial version to "initail" in order to identify the version with this more meaningful name.
Fred wants to access Joe's initial version of index.html. So he uses URL http://aws/ags/index.html to get the contents of the resource and notices he does get the right version, because it was selected by the default workspace. That is, when Fred accessed URL http://aws/ags/index.html, he did so without specifying a workspace. So the default workspace was used, and the default workspace always uses "latest" as its version selection rule. But Fred wants to be more cautions. He wants to be sure that he continues to get version labeled "initial", even if the latest version changes as the result of new changes Joe may check in. So Fred creates a workspace called "initialws", and sets the version selection rule to be the revision labeled "initial". Then Fred always access index.html with its URL and the initialws workspace to be sure he gets the specific version he needs.
A week later, a number of developers have noticed that index.html is missing both important references to their pages as well as hot images for navigation. They send email to Joe specifying their new requirements. Joe now wants to make changes to index.html and create a new version. He wants to retain the old revision, just in case the requirements he was given were incorrect and need to be backed out, and to allow developers using the old revision to continue their work. To do this, Joe uses the WB to check out index.html and create a new working resource. Joe can now access the working resource because working resources are always visible from the workspace in which they were checked out.
As before, Joe uses the WB and HTML editor to GET the working resource and PUT updates. Each PUT replaces the contents of the working resource with changes made by the HTML editor, no new revision is created. When Joe is finished making edits to support the new requirements, he checks the working resource back in, making a new revision.
John was assigned to write a high level marketing document, ags.html, that provided an overall description of the AGS application. Since most changes to this document have no effect on the rest of AGS, John decides to allow revisions of ags.html to be editable. This is so simple spelling and grammar errors can be fixed without requiring the creation of a new revision. John still wants to create revisions whenever some significant new feature is added to AGS so the old descriptions are available to customers who don't upgrade.
John creates resource ags.html, edits it a number of times, and then checks it in to create a versioned resource. But in this case, John checks in ags.html as a mutable revision so he can fix spelling mistakes, etc. by simply doing a PUT to the mutable revision, just like a non-versioned resource or working revision.
Later on, a new feature is added and John checks out ags.html to create a new revision, makes his edits and checks it back in, again as a mutable revision. In this case, the descendent is related to the ancestor by a was-derived-from relationship to indicate that ancestor revision might change.
Six months later, there have been a number of complaints about ags.html presenting misleading product information that has resulted in unhappy customers. There's even talk of law suits (read money-making opportunities). So John hurriedly updates ags.html and checks in the new version as immutable so that in case there is a suit, he can prove that customers had access to his updated version. Now any changes can be made by creating new immutable revisions without ever worrying about loosing old version.
A year later, things have cooled down, and John decides its OK to allow mutable revisions again. On his last change he checked ags.html in as a mutable revision allowing subsequent changes to be done without creating new versions. At the same time, the revision history of the immutable revisions is preserved just in case that pesky customer re-appears.
Two weeks later, there is a major redesign of AGS that results in a lot of changes to index.html. Again, Joe checks out the resource creating a new working resource. But it is taking Joe a long time to finish all the edits, and in the meantime, graphics artist Jane wants to update index.html with references to the new images that resulted from the AGS redesign. Jane attempts to check out index.html, but WB informs her that Joe already has it checked out and refuses the request. She checks with Joe, and since they are both working on different aspects of index.html, Joe feels it would be fine for Jane to do her work in parallel with his, and then he will merge her changes with his to finish the required updates. Jane creates a new activity called "images_updates", uses it to set the activity of her workspace, and again attempts the checkout. This time the checkout succeeds, and a new working resource is created for index.html in the images_updates activity. Now any changes that Jane makes to images.html are completely independent of changes Joe makes to the same resource, but in a different activity. Note that Joe did not create an activity when he checked out index.html. Instead, the default activity "mainline" was used. Jane couldn't checkout index.html without specifying a different activity because a resource can only be checked out once in a given activity.
After making her edits, she checks index.html back in, which creates a new revision in the images_updates activity. There is still no need for Joe or Jane to lock index.html as they are working on different working resources in different workspaces. Again, either of them may lock their working resource to ensure no one else can change it.
Project management practice dictates that at various times during the development project, usually every few days or at specific project milestones, the updates from any parallel activities should be merged in order to integrate the changes and produce instances of the products suitable for testing. This avoids the risk of revisions of shared resources diverging wildly, and thereby decreases the likelihood of difficult or inaccurate merges. It also encourages communication within the development organization and avoids "big-bang" integration points late in the development cycle. This enhances the stability of the products and helps ensure a deterministic, controllable development process. It also allows early product testing and better feedback to developers.
Joe has finally finished his changes to image.html, and is ready to incorporate the changes from Janes images_update activity to get the new images. Before doing so, Joe checks his updates into revision "r0.2" so if he does something wrong when doing the merge, he can recover and try again. Now Joe specifies in his workspace that he wishes to merge the "image_updates" activity into his workspace. He then can obtain a conflict report from his workspace which indicates that the resource index.html requires a merge. He then issues a merge request for index.html. This checks out the resource in the mainline activity (the activity in Joe's workspace), and registers a merge from the latest revision in the image_updates activity to the working resource. This working resource now has two predecessors, r0.2 and the image_updates revision. Joe then uses the differencing capability in his HTML editor to find the differences between his revision and Janes, and to apply Jane's changes as appropriate.
The HTML editor Joe uses is WebDAV versioning aware, and does a 3-way merge by accesses
the closest common ancestor in the revision history in order to help with the merge
process. Joe notices that most of Janes changes do not conflict with his as they are in
different places in the resource, but there are a number of places where he added new
functions that do not have images as Jane didn't know they were there. He notes these and
either fixes them himself, or sends email to Jane so she can fix them in another version.
Once the changes are complete, Joe checks in the merged version. Jane is free to continue
making updates in her image_updates activity, and these changes can be merged in again
There is still no need for Joe or Jane to lock index.html as they are both working on different working resources. Again, either of them may lock their working resource to ensure no one else can change it.
At some point, enough of the work products of the AGS application are sufficiently complete and stable that AWS wants to distribute an alpha release. To do this, Joe uses WB to create a configuration named "alphaRelease" that will contain a consistent set of compatible work products. This configuration will contain all revisions currently selected by Joe's workspace. If any working resources exist in Joe's workspace, the request to create a configuration fails, with an error message indicating that the failure is due to the presence of checked-out resources in Joe's workspace.
When Joe is happy with the state of the his workspace (including the mainline activity), he can create a "configuration" (see below). When Jane is ready to see this new state, she would modify the version selection rules of her workspace to select this new configuration. Any conflicts between this new configuration and her current activity requiring merges would be noted in the "conflicts" report of her workspace, which Jane could then resolve with the "merge" operation.
This section defines the goals addressed by the protocol to support versioning, parallel development, and configuration management. Each goal is followed by a short description of its rationale to aid in understanding the goal, and to provide motivation for why it was included.
Some of these operations come from [WEBDAV-GOALS],
section 18.104.22.168. Not all of the operations in that section are replicated here; some
of them (e.g., locking) fall out naturally from the fact that a revision is a resource.
The protocol must find some balance between allowing versioning servers to adopt whatever policies they wish with regard to these operations and enforcing enough uniformity to keep client implementations simple and interoperable.
Versioning in the context of the world-wide web offers a variety of benefits:
It provides infrastructure for efficient and controlled management of large evolving web sites. Modern configuration management systems are built on some form of repository that can track the revision history of individual resources, and provide the higher-level tools to manage those saved versions. Basic versioning capabilities are required to support such systems.
It allows parallel development and update of single resources. Since versioning systems register change by creating new objects, they enable simultaneous write access by allowing the creation of variant versions. Many also provide merge support to ease the reverse operation.
It provides a framework for coordinating changes to resources. While specifics vary, most systems provide some method of controlling or tracking access to enable collaborative resource development.
It allows browsing through past and alternative versions of a resource. Frequently the modification and authorship history of a resource is critical information in itself.
It provides stable names that can support externally stored links for annotation and link-server support. Both annotation and link servers frequently need to store stable references to portions of resources that are not under their direct control. By providing stable states of resources, version control systems allow not only stable pointers into those resources, but also well-defined methods to determine the relationships of those states of a resource.
It allows explicit semantic representation of single resources with multiple states. A versioning system directly represents the fact that a resource has an explicit history, and a persistent identity across the various states it has had during the course of that history.
These non-goals enumerate functionality which the working group has explicitly agreed to exclude from this document; they are documented here for explanatory purposes.
To be written. It is likely that implementing features to meet the goals described here will present few or no new security risks beyond those of base DAV. One possible exception is that it may become more difficult to hide the contents of a resource when there may exist other versions with different access control lists.
[WEBDAV]Y.Y. Goland, E.J. Whitehead, Jr., A. Faizi, S.R. Carter, D. Jensen,
"Extensions for Distributed Authoring on the World Wide Web -- WEBDAV",
[WEBDAV-GOALS] J. Slein, F. Vitali, J. Whitehead, D. Durand, "Requirements for a Distributed Authoring and Versioning Protocol for the World Wide Web", RFC-2291. February 1998.
[WEBDAV-ACP] J. Slein, J. Davis, A. Babich, J. Whitehead, "WebDAV Advanced Collections Protocol", Internet-Draft draft-ietf-webdav-collection-protocol-02.txt. Nov., 1998.
[DASL] S. Reddy, D. Jensen, S. Reddy, R. Henderson, J. Davis, A. Babich, "DAV Searching & Locating", Internet-Draft draft-reddy-dasl-protocol-04.txt. Nov., 1998.
[BONSAI] Mozilla.org, http://www.mozilla.org/bonsai.html