Point/Counterpoint: Software as a service

A few of us at InfoWorld latch onto the latest trend and run with it. Others retain their disbelief, even as enterprise IT spends big bucks on new and exciting technology. The hot trend of the moment, software as a service, has stirred energetic debate inside InfoWorld. Our leading enthusiast (some might say “patsy”) for the latest and greatest, Executive Editor at Large Eric Knorr, and our foremost skeptic (some might say “young fogey”), Test Center Techincal Director P.J. Connolly, have been arguing about the software-as-a-service trend, led by Salesforce.com, ever since Knorr wrote “The End of IT as We Know It?”. Their ongoing argument has generated some light as well as heat, so we decided to get it on the record for the (some might say “dubious”) benefit of InfoWorld readers.

EK: My basic proposition is this: Salesforce.com-style hosted services will replace in-house, server-based applications sooner than most people think. The most compelling technical reason is that Internet protocols have already taken over the core of the enterprise, so most enterprise apps are being delivered through the browser already. Who hosts the app has little or no impact on the end-user experience. But the business issues are what will really drive this: You can’t beat zero deployment and maintenance combined with zero up-front licensing costs. Granted, establishing trust for large-scale deployments is a challenge for the hosted model. As the trusted brands emerge, though, there’s no stopping this trend. These days, can anyone predict where their business will be 18 months from now when all the checks have been written for software licensing and new server hardware, and when some huge enterprise software rollout is finally complete? Why not just pay as you go, pay for what you need?

PJC: Eric, I’m all for cutting the ties to legacy applications, but there’s a fair amount of reality that has to be checked before IT departments can dump all of their in-house applications. Funny thing is, I was going to mention Salesforce as an example of the kind of application that’s easy enough to digest, because sales is sales. It doesn’t really matter if you’re buying big ones and selling little ones, or if you make big ones and sell big ones; inventory is still inventory and commissions are still commissions. The problem is what happens behind the Web browser: There’s still a lot of mainframe screen-scraping going on, and that’s where many of these one-size-fits-all schemes fall apart. Another thing: Who told you there’s zero deployment cost with hosted services? There’s always something that takes more bodies or time, even if it ends up being a soft cost.

EK: OK, I’ll drop the jargon. By “zero deployment” I mean no upfront cost in hardware or software, which you must admit lifts a burden from IT managers who would otherwise have to justify the capital investment. Your main point, though, if I understand it correctly, is that only some kinds of apps — such as CRM or inventory — lend themselves to outsourcing, because there’s little need to customize, whereas many legacy and/or in-house apps are specific to a vertical or to a company’s individual business processes. I agree that these would be last on the list to be delivered as a service. I’m not suggesting a big bang in which you’d outsource everything at once. In general, if homegrown legacy apps work, there’s no need to replace them unless the maintenance costs become unbearable. If you’re in the market for a commercially licensed, server-based software solution, though, you should also consider software as a service. In fact, I submit to you that the lower cost of entry makes the service more attractive than licensed software, provided of course that the service has the features you need.

PJC: In general, I have to agree with that statement, but the model has its weak points. One problem with a thin-client model is that it’s usually predicated on easy access to “fat pipes” — the fast network connections that we take for granted in the office. Decent network access is still an issue for people in the field, and for many businesses these folks are the moneymakers. Security is still an issue for me: Why is an application that relies upon a fundamentally flawed tool — that’s Internet Explorer if you haven’t figured it out already — better than a fat client that isn’t a target for every pimply faced youth who thinks he’s the next king of the VXers? But back to your point: Of course IT managers should be looking at hosted solutions, but they have to be looked at as a means to an end, not the end itself.

EK: Technology as an end in itself? Nah, that’s for you guys in the lab coats. Look, I actually think your security argument is making my point: IE is a problem for most Web apps, not just the subset we’re calling software as a service. And when it comes to enterprise software, homegrown or shrink-wrapped, Web apps rule because deployment is so much easier. The security problem isn’t going to be any worse with software as a service. I guess some applications should remain client/ser-ver based, but even the ERP vendors have gone to browser-based clients. As for network access, your folks in the field will have that narrow-pipe problem whether they’re logging into the company VPN or into a hosted service. But we haven’t really touched on one of the most compelling aspects of the technology: The ability for customers to customize hosted applications “safely,” in a controlled remote environment, so that those modifications don’t break new versions of the hosted application. That’s been a huge stumbling block for enterprise apps. Hosted services like Salesforce.com’s sforce are finally solving the problem.

PJC: I don’t own a lab coat. What I do own is an understanding of the problem, and saying IE is a problem for most Web apps is like saying that carbon monoxide is a problem for most animals that breathe. You’re saying that because deployment is so easy, I should ignore the security problems inherent in the browser that overwhelmingly dominates the installed base. Nope, I’m not buying that fish anytime soon. I’d feel a lot better about Web-based apps if they were truly browser-agnostic, and the best ones are just that. At least — to overuse an example, but it’s one I’m most familiar with — Salesforce.com lets you use Netscape Navigator with the service. It’s a step in the right direction, and aspiring hosted-service vendors would do well to avoid platform-specific gimmickry. Finally, what you describe is not customization; it’s picking predefined options. That’s about as custom as the interior choices at your car dealer.

EK: I’ll just leave alone the implication that you’re backing into a defense of client/server. A few InfoWorld readers might disagree with you on that. Funny you should mention car dealers, though. Chad Dickerson recently wrote a column about the hazards of customization lately. His thinks the desire to have things exactly your way underlies a host of IT problems. Plus, I would argue that sforce goes way beyond picking features from a menu. It’s really a development platform, although what you can develop may be limited compared to what you can achieve with, say, PHP spaghetti code. I think a better refutation of what I’m saying might be that Salesforce.com and sforce, along with CrownPeak, Grand Central, and a handful of others, pretty much stand alone as hosted app dev platforms. I’m convinced that they will be joined by others, however. It’s going to be part of the hosted model.

PJC: There was a project I worked on more than a dozen years ago in which the customization got so out of hand that I still wake up screaming. I won’t even mention liver damage. But to your point that sforce should be considered as a development platform, that’s neat. Really. The problem with applications that grow into development platforms, though, is that customers wind up writing slop code that can’t be fixed in-house. Look at all the orphaned Lotus Notes applications out there and you’ll see what sforce could look like in five years if its guardians aren’t careful. Of course, hosted application development is integral to the hosted-service model. I can’t imagine it working any other way. I also can’t imagine hosted services putting in-house developers out of business. It just isn’t going to happen.