We've been working in the Paypal sandbox and Authorize.net test account lately and neither really feel like the real thing. Paypal's sandbox is slow as molasses, among other very irritating foibles. Testers need to login separately in the sandbox, while this really should be done through code and be transparent to the users testing the frontend of a web application. The fake credit card numbers provided with the sandbox also most often don't work properly. It feels like pulling teeth to test an end-to-end transaction.

At the end of the day we decided Paypal was too amateurish in lots of ways. Their Payflow gateway has a dismal service reputation since they acquired it from Verisign so we're going with the whole Authorize.net + merchant account shebang instead. Our tests there look better though, here again, you don't get the feeling the test account behaves fully like the real thing. Also, despite being one of the leading payment gateways, Authorize.net doesn't seem to have a health status page a la trust.salesforce.com.

Web service vendors and API providers, if you're providing a sandbox, please, make it work like your full-fledged product, and don't run it off an old Pentium in your basement.

I haven't heard back from Yahoo about their intent with Pipes so I'll assume they're just screwing around and having fun (which I guess is a somewhat valid statement about Yahoo as a whole - can Semel close the door after him when he leaves?). To see applications that actually show some business potential, read From Web 2.0 to Work 2.0 and look at this RSSBus+Proto demo.

The problem with business mashups is to find data sources that are clean and specific enough. We manage lead generation in SFDC (did I mention how much I love Salesforce.com since we started using it six months ago?) and use data from sources such as Hoover's, LinkedIn or Google queries, but that stuff is too unstructured or imprecise to support automation. Short of human oversight, you'll often end up with Garbage In Garbage Out. Garbage leads means your reps will ignore your SFA and you don't want that. So we have custom links in SFDC that pop up queries from various sources, and the rest is grunt work.

Proto's expense demo works because the data is already structured and it's part of any corporate workflow to tag and review expenses one entry at a time (aka "stop entertaining your girlfriends at the company's expense"). Expense management was one of the first applications on the intranets emerging a decade ago. I'd like to set up mashups based on things like keyword density to support editorial workflow or competitive research, but so far I haven't seen very convincing tools so far. From an automation perspective, screen-scraping is not going away anytime soon, and again, there's only so much you can automate without actual AI semantic understanding

Here's something that business web app providers should do that would be a boon right here right now: enable web queries directly from Excel as opposed to forcing users to do time-consuming manual export jobs. Excel has been a hybrid web/desktop client for years but we're still stuck downloading CSV files which, if you think about it, is a pretty sad thing to do in this day and age. I have to spend way too much time shuffling and massaging data around to get the information I need.

A broader point is that relying solely on browser-based authentication to grant access to online apps is a huge lock-in. It's shutting out desktop apps that don't know how to navigate within HTML interfaces to get to the data, just like it's shutting out, say, WiFi phones from making Skype calls from the local Starbucks. I'm all for paid online services, but please don't serve them within a web ghetto.

Like many people, I find Yahoo Pipes to be a bold and exciting move. What I don't quite get is what audience Yahoo is eventually planning to go after. I guess I don't see how you make money out of this as a free service. This stuff is middleware, how do you slap ads around it? I wish Yahoo and Google had the same maturity Amazon displays with S3. I need to see a clear intent and pricing before I'm going to consider embedding such tools in our business processes.

I hope my post will get an answer:

"Is this a souped up RSS reader for power users? Or is this a middleware platform offering that companies will be able to rely on to pipe in data from partners? In other words, is Yahoo Pipes meant to be a Bloglines or an Amazon S3? Is Pipes going to come with an SLA, scheduled downtimes and all the stuff you come to expect from SaaS?

In the first case it's just a nice toy to play with. In the second case it might be a tool we want to use in my company to integrate data from third-parties with our own and deliver it to our readers. Right now we have to write custom php code to parse feeds and pump third-party content into our CMS templates. I wonder whether Pipes could empower "business analyst" types who don't know how to code php but know what kind of data they want to put together for a specific audience."