Some time ago I wrote about how the "programmable web" has really turned out to mean the queryable web, because real transactional APIs have not been readily available.
Public web services (with a few exceptions, mainly around payments and online file storage) are a way of retrieving state. Using web services to execute useful "write" transactions into other systems has turned out to be the business of enterprise SOA, not late-night hacker camps.
The concern, of course, is over users "getting it wrong" when insufficient semantic data is available. Although that's a bogus argument (many industries and services have solid semantic models that are readily accessible from web service descriptions), most folks seem to be waiting for the legendary semantic web to come along.
But "most" isn't "all," and I was glad to see O'Reilly write about Public Domain Reprints, a transactional mashup that submits books to print-on-demand services where you can later order a copy.
Although Public Domain Reprints is just formatting and "preparing" the book -- you still need to order a copy yourself from the service of your choice -- the mere adding of an item into a print-on-demand service's catalog is fairly exciting given how rare these "commit data"-type open web integrations (mashups) are.
At this stage, it doesn't really matter what sort of integration mechanism is used -- it needn't be a web service; it could be emailing a PDF into a dropbox, or simulating a file upload from a web form. Obviously, having a "strong" interface (including some kind federated authentication, so that a mashup can act on my behalf using a restricted token and not my password) will make more services feasible.
I continue to hope that we will see more services that don't just pull down a bunch of data and create a nice report, but actually reach out to services I use and improve my life by doing a little bit of work for me.