Improved usage flow of faces.train

As part of our continuous effort to improve API performance and flow, we are pushing a small change in the faces.train API method.

Faces.train is a potentially long operation.  When used with long uids list, or with  friends@facebook keyword, it may create a work queue of thousands of photos.
There is no point waiting for that kind of workload with synchronous method call. Most of our developers, are already using training their uids with asynchronous callback_url call.

So, we decided that faces.train for more than one UID, will only be available via asynchronous call - when the callback_url is provided.
When used with more than one UID, a callback_url must be provided (if not, a SYNCHRONOUS_REQUEST_TOO_BIG will be returned).
Calling the method with single UID is allowed also without a callback_url.

The callback_url communication is also slightly changed:
Instead of POSTing the entire result of all UIDS to the callback_url one all of them is completed, we will POST partial response to the callback_url, for each UID that his training is complete.

We believe those changes will improve the training flow, and will scale better when developers will use it for more and more uids.

For more information, check our http://developers.face.com/docs/api/faces-train/.

-- the face.com team