On Fri, Aug 6, 2010 at 8:39 PM, Nicholas P Cipriano <[email protected]> wrote: > 1) I would consider making this an "all or nothing" operation. For example, > You could program the server to step through each element in the list to > ensure that all of the change requests will succeed. As soon as you find an > error, return that error to the client using the format described in XEP-0045 > (do not commit any of the changes) and exit the operation. If all changes > will succeed, then I would iterate through the list again and commit them all.
Yes, this is certainly one of the ways I've considered. The biggest drawback (besides somewhat more complex code on the server) would be that the user would not be informed on the first round of errors that will happen when a resubmits. Like in my example, if he only get the not-allowed error, he will fix that and resubmit, only to get the "conflict". /niklas
