Currently, when the InputType=Text and AllowCustomEntry is set to true, the AutoCompleteBox does not persist the value of the entry when an item from the dropdown is chosen. 

Instead of the control expanding vertically to accomodate all entries, they should be positioned on a single line.

Based on customer feedback: there should be a property that limits the maximum height of the input of AutoCompleteBox.

Once that height is reached, a scrollbar should appear.

When typing in a text control it would be nice to have auto-complete based on some predefined values.  

In my example we have a list of names setup against a handle like @johnsmith.

If typing in the textbox when a person types a @name it would check against the available names and validate, if invalid they would be presented a list to pick the correct name (like how spell check works)

This was brought up in a support ticket, notes below:

Do have the following in mind:

* implementing the simplest case e.g. @mentions will simply not do. there are tons of other completes that can be done e.g. #hashes, :emoji and so on
* developers will need to verify that tokens are indeed valid and what to do with invalid tokens e.g. if I type @nosuchuser and there isn't such user, should the token be left literal or should it be processed like the rest?
* there should be a way to specify the data source location for those completes...

Perhaps more.

But the most important part was the emphasis on the token validation and data source.

Apart from that the logic is fairly simple: a textbox with events for keypress. As soon as a trigger key is pressed the autocomplete part pops and the user either selects a value from a list or types one. The "rich" value is saved in a hidden field. If the value happens to be a valid one, it has special formatting (because at the end we are submitting the rich value).

Instead of having the property to select single or multiple it would be more productive to have a property called SelectCount and we can enter the amount of listings to be selected. If its a single I select 1... for multiple I select how many I need. No more need for the Single/Multiple property except for backward compatibility.

I'm using ASP.Net AJAX RadAutoCompleteBox contrl. I can go throught the filtered items with 'up' and 'down' keys, but i want to select an item with 'TAB' key, is that possible? 

In your page http://www.telerik.com/help/aspnet-ajax/autocompletebox-onclientrequestfailed.html you only have one message to show to the user, why not return de http status code?

In this moment the workaround suggested by your team is:
<script>
    Telerik.Web.UI.WebServiceLoader.prototype._onWebServiceError = function (error, context)
    {
        var errorEventArgs = new Telerik.Web.UI.WebServiceLoaderErrorEventArgs(error.get_statusCode().toString(), context);
        this._raiseEvent("loadingError", errorEventArgs);
    }
 
    function OnClientRequestFailed(sender, args) {
        console.log(args.get_errorMessage());
    }
</script>

https://demos.telerik.com/aspnet-ajax/autocompletebox/examples/accessibility-and-internationalization/wai-aria-support/defaultcs.aspx

While JAWS will announce all of the items in the list as results are first read, it only says "Blank" as you arrow through the available options. Tabbing out of the control and back in, also does not read the selected content.

This is basically the opposite of what happens when ARIA support is turned off.

 https://demos.telerik.com/aspnet-ajax/autocompletebox/examples/default/defaultcs.aspx

In that case, there is no announcement when results come back, but JAWS will read each individual item as you arrow through the available choices. Tabbing out of the control and back in still does not cause JAWS to read the selected item. 

This makes it seem like the ARIA version of the autocomplete box is missing either the appropriate ARIA roles or some other piece of rendered markup that is present in the non-ARIA-enabled version.

 

Ideally, when working with screen readers,  we would like to see something like the following behaviors:

* The number of items are announced when results are loaded, including empty results

* Each item is announced as the focus is moved through the list

* The selected item is announced when picked or whenever focus returns to the field.