&xep0050; defines a mechanism to enhance data forms creating dialog wizards with actions that guide the user through pages, where each page can depend on the input from previous pages. The following list consists of a comparison of the differences between Ad-hoc commands and Dynamic Forms, and why this extension is not based on the concepts defined in XEP-0050:

• The main point of this extension is to make the current form dynamic. Pages in a wizard, as defined in ad-hoc commands, are still static requiring page navigation for the dialog to be updated.

• There might be a desire to make individual pages inside a wizard dynamic by themselves. In this case, this extension might be used to extend Ad-hoc commands with dynamic pages inside wizards.

• Wizards order their pages linearly, in a one-dimensional array of pages. Navigation is only performed using prev and next actions, and completed with the execute or complete actions. This might work sufficiently well where page output only is based on input from the previous page. But as soon as the input depends on more pages, navigation becomes unnecessarily cumbersome. For example a dialog with country, region, city, area, street, building, apartment. To change the country parameter when you are on the apartment page, requires you to back through many pages, while in the dynamic form you just change the parameter directly, since all are visible in the same form.

  The problem gets worse if this is a normal behavior for a form. Say a dialog with a post carrier field, followed by country, city, office, etc. When the user comes to office and notices that the selected carrier has no office close to you, you change carrier, but don't want to change country and city, if possible.

• The pages in a wizard do not allow for spontaneous or immediate feedback to the user. Such immediate feedback is of great importance in user-friendly forms, where complex input is required. Consider the following simple example:

  Imagine a dialog used for configuring a TCP port scanner to search for specific devices in a network. It might take five parameters: IP range, port range, number of threads, connection timeout (ms) and number of tries before failing. However, the operator wants to know how much time the operation will take (which is #IP-addresses * #port numbers * Connection Timeout * Nr Tries / #Threads). To see this in a separate page and then have to go back to the different pages, forwarding to the time estimate calculation page, etc., is a very cumbersome process. Presenting all this information on a single page using Dynamic Forms is creates a much better and richer user experience. Changing the value of any of the five parameters will directly update a text parameter presenting the total expected time of the operation.

• Wizards as defined in Ad-hoc commands are hard-wired to commands as such. Navigation actions are defined outside of the scope of the form itself. This extensions defines Dynamic Forms as an extension of the Data Form concept. All navigation, post-back fields, etc., are defined within the actual form itself. This makes it re-usable everywhere data forms are used, including pages inside an Ad-Hoc command wizard.

• Ad-hoc command wizards include the feature to return notes including information, warnings or error information to the user. This information however, is associated with the current page in the wizard itself. Dynamic Forms contains a means to attach error information directly to individual fields, making feedback more precise. Together with the server post-back feature immediate server-side validation during user input is possible.

• Dynamic Forms contains a feature that ad-hoc command wizards do not: Read-only fields. These differs from other text fields in that they are rendered as normal controls, except input is read-only. Together with the server post-back feature this allows the server to enable and disable parameters depending on user input on the same page. An example can be a checkbox that enabled another field if checked, as a security measure.

• Dynamic Forms defines the concept of an undefined field value. These fields are presented with the available value (if any), but greyed out, signaling to the user that the value is not well defined for some reason. Perhaps it is only one of many values held my multiple objects in the case when editing multiple objects at once. Undefined values are not sent back to the server when the form is posted back. When the user edits such a field, the undefined flag is cleared and the field is presented as a normal control. The main purpose of this flag is to allow for editing multiple objects at once or for editing control forms where current states are not known at the moment if creating the form.

• Dynamic Forms supports asynchronous updates of the form, where the server can push changes to the form not resulting from user input. This is a powerful feature that allows the server to update a form being edited by the user to reflect changes on the server. Consider the following examples:

  Consider a control form containing control fields on a remote device. At the time of displaying the form, the current states of some fields might be unknown. So the fields are marked perhaps with some default values, but with the undefined flag set for the corresponding fields. At the time of creating the form, a parallel request is made to the device by the form server, requesting information about current states of the device. When these are received by the server, it issues an update of the form to the client with the newly received and current field values. The request for values might take some time, so using this mechanism provides a form to the user quickly, clearly indicating what is missing, and then complements the form when data is available.

  Another example might be a dialog showing contents on the form server in a multi-user environment where updates to the contents is made. An example can be a file system. If changes to the contents is made by another user, the server has the possibility to update any current forms to reflect changes made. This decreases the possibility of inconsistencies in the system, and at the same time increases the user-friendliness of the end-user experience of the application.

Post-back fields are fields that require the client to post the form to the server after being edited. This permits the server to perform server-side validation on the field, provide immediate user feed-back and change the form according to user input.

Post-back fields are declared as normal fields in a form, except they also have the xdd:postBack flag present in the declaration, as is shown in the following example:

Note: A system maintaining multiple dynamic forms open at the same time needs to maintain control of available open forms. This can be done using a hidden session variable, as is shown in the example above. How this is done however, is implementation specific.

A server post-back is performed by sending an IQ set stanza with a formPostBack child element containing the current state of the form. The type of the form must be submit. The client should also provide the current user language in a xml:lang attribute, if available.

It is important to note that this server post-back is part of editing the form, and must not be treated by the server as a final submission of the data form.

As a response to a successful form post-back, the server returns a formPostBackResponse element with a result code of OK containing a new data form, as is shown in the following example:

In this example, the server adds a new parameter to the form, containing options that depend on the value of the first parameter.

Note: Server post-back should be made after the user is done editing a post-back field, but before actually navigating to the next field. It should not be done while the user is editing the field. For check boxes or list boxes, it is easy for the application to decide when the field has been edited, since the application can react to the controls click event. But for text edit boxes, you cannot perform a server post-back only just because the text property of the control has changed. You need to wait until the user leaves the control or something similar. However, as new fields can be added to the form, the client should wait for the response before deciding which control is the next control to go to.

Read-only fields should be laid out on the form just as a similar but editable field would, except editing should be disabled. Together with post-back fields, this option allows the form to enable/disable fields depending on user input. Read-only fields are declared as normal fields in a form, except they also have the xdd:readOnly flag present in the declaration, as is shown in the following example:

Note: Using this flag is different from the field type fixed defined in Data Forms. Fields of type fixed can also be used to display read-only content. However, this is done as static text, and not as a disabled control, specific for the type of content corresponding to the data in question. Using the readOnly flag instead, gives greater flexibility when it comes to presentation, as well as permitting the form server to enable and disable controls during the lifetime of the form.

Note 2: Make sure to check the implementation note Merging Client-Side Values for information on how to merge updates received from the server with current input made by the client.

There are many cases where you want to flag a control as having an undefined value or an uncertain value, instead of simply a missing value. This permits the form client to display the control in a specific way (for instance greyed), and omit the field when submitting the form back to the server, to avoid errors.

This technique can be used for instance, when editing the properties of multiple objects at the same time. Attributes that are equal among the objects can be reported as normal fields, while attributes that have different values among the objects, can be reported as having a value that is not the same everywhere. The client can then show the corresponding control greyed and omitting it in submissions of the form. If the user edits the control however, the form client can remove the flag and render the control normally. Submitting the now well-defined field value will set the corresponding attribute in all objects to the same new value. Omitting undefined values makes sure that the corresponding attributes are left as-is.

Fields with undefined or uncertain values are declared as normal fields in a form, except they also have the xdd:notSame flag present in the declaration, as is shown in the following example:

In the above example communication properties of a set of objects are edited. The form contains two fields, one address field, which will contain unique values for each individual object and therefore be flagged with the xdd:notSame flag, and a second baud rate field. This second field will probably contain the same value, if devices are connected to the same bus. Therefore, the field value is not flagged with the xdd:notSame flag.

Note: If the xdd:notSame flag is available in a field, the field must not be flagged as being required. They must also be omitted in any form submissions unless they have been edited first. Furthermore, any field submitted in a post-back as described in this document, must not be returned in the post-back response having the xdd:notSame flag.

Notice also that there's a difference between a missing value, i.e. a field without a value element defined, and a field with a value defined, but with the xdd:notSame flag present. In the latter example, the value might represent the value of one of the objects in a set of objects being edited.

A field can be flagged with an error message using the xdd:error element. Together with the post-back feature this allows for more advanced server-side validation of field values, as is shown in the following example.

In the above example, the user can enter a script expression into a text box using a server-specific scripting language. There's no way for the client to validate the script expression. However, by flagging the field with xdd:postBack the client posts the form back to the server when the user is finished writing the expression, and the server can validate it. In the example, the user has made an error, and the server returns the field, with an xdd:error flag, containing an error message that can be displayed to the user in an appropriate way.

Note: If a field is flagged with an error, and the user starts editing it, the client should remove the error flag of the field. The field can be flagged again with a new error flag during the next post-back.

The Data Forms XEP provides a mechanism to cancel forms, by submitting the form using type='cancel'. The Data Forms XEP stipulates that fields should not be included when using the form type cancel. However dynamic forms, i.e. forms containing post-back fields, will need the fields, since any dynamic form session will be identified using hidden fields in the form. So, to cancel a dynamic form, the xdd:cancel element with the entire submitted form should be sent to the form server using an IQ set stanza, as is shown in the following example:

After receiving the cancel request the form server returns a cancel response having a result code of OK if the form was found (and therefore cancelled), or NotFound if the form was not found. The following example shows an OK response:

Note: If cancelling a dynamic form using the approach described in this document, there's no need to also submit a cancel form as defined in the Data Forms XEP. The form server automatically makes sure the form is cancelled in all instances on the form server.

It might happen that the form server does not find the dynamic form posted by the form client during a post back. Reasons for this can be that the form does not include a post-back field, or that a form session timeout has occurred and the form server has discarded the dynamic form to avoid memory leaks. Regardless of the reason, the form server responds using a NotFound result in the response, when the client posts a form that is not found back.

Note: Form post-back must only be performed on forms containing post-back fields. The form server is not required to maintain dynamic form sessions for forms that lack post-back fields.

If another error occurs during post-back, the form server can inform the client about this, using the OtherError result code, and include an error element to describe the error, as is shown in the following example:

The server may send asynchronous updates to open forms on the client. This can be done if the server detects changes that it wants to inform the client about. Changes are made by sending a message stanza including a formUpdated element which in turn contains the new updated form.

Examples can include a control form showing control parameters. While the server is trying to retrieve the current values it presents the control form with undefined values, and later when values are received by the server, it sends an update to the client with actual values.

Another example can include a dialog containing information on items on the server in a multi-user environment (for instance a file system). Changes made by users can be displayed in open dialogs to other users as they change.

The client may have more than one form open at any given time. It might also be so that the form has been closed prior to receiving or handling the update message, and is therefore no longer visible. To be able to identify to which form the update corresponds, the formUpdated element is required to include a sessionVariable attribute in which it identifies a unique identifying field in the form. When the client receives the update, it goes through all forms it has open. If a form has a field variable with a corresponding name, and the field variable has a value equal to the value in the updated form, the form should be updated by the contents of the message. If no form is found, the update is simply ignored. If multiple forms are found, all should be updated.

Note: Make sure to check the implementation note Merging Client-Side Values for information on how to merge updates received from the server with current input made by the client.

For a form server to maintain the status of the dynamic form, it will probably need to publish state or session information using hidden fields in the dynamic form. It's important that form clients be aware that such hidden fields are available and must always return them in all submissions of the form to the server.

A form client that supports dynamic forms and opens a dynamic form, i.e. containing parameters requiring post-back, must call the specific cancel method described in this document to cancel the form if not submitted. This to let the form server to release any dynamic form session resources it maintains.

Note: When submitting a dynamic form the normal way, any dynamic form session resources are also automatically released.

The form server must be aware that some form clients do not support dynamic forms. This in turn implies that form clients may not call the correct cancel method to cancel a dynamic form. To protect the form server from memory leaks, it must include a session timeout, and release any dynamic form session resources if no activity has been made during the corresponding time period. If the client would perform a post-back after the timeout period, a NotFound error message must be returned, to show the corresponding dynamic form session no longer exists, and therefore could not be found.

For normal operations, a dynamic form session timeout of 15 minutes is sufficient.

When receiving asynchronous form updates from the server, or when performing a server post-back of a form, it is important to know how to merge responses from the server with the current form being displayed to the user. As the operation is asynchronous, and since user input is quick, the user might have input things not known to the server and therefore not available in form updates. Also, fields not marked for post-back might not have been reported at all to the server, and therefore, the client is the only one that knows what the user has entered into these fields.

So, when receiving form updates, either asynchronously, or as part of a server post-back response, the client needs to merge the updated form, with the current form. The following rules must be applied. Here, the updated form represents the form in the recent message from the server, the current form represents the form currently being edited by the user and the resulting form represents the result of the merger of the updated form and the current form.

• New fields in the updated form not available in the current form, are added as-is to the resulting form.

• Fields not available in the updated form but available in the current form, must be removed from the resulting form regardless if user input is available. Any such user input is lost.

• The order of fields in the resulting form must be the same as the order of fields in the updated form.

• Fields available in both the updated form and the current form are handled depending on if the user has entered values in the corresponding field in the current form or not:

  • If the user has not edited the value in the corresponding field, the value from the updated form is used.
  • If the user has edited the value the corresponding field, the value from the current form is used.
  • If the user has edited the value of the corresponding field in the current form, but the value is the same as the value available in the updated form, the flag stating that user input has occurred in the field can be cleared.
  
  

• If a field in the updated form is flagged as undefined, but the current form has an edited value, the form in the resulting form must not be marked as undefined.

• All other properties for fields must be taken from the updated form and copied to the resulting form.

How the above merger is made is implementation specific. One simple implementation can simply be taking the updated form, adding any client-side values to it (i.e. values edited in the current form) perhaps removing any undefined value flags, and then use the result as a model for the resulting form.