Introducing pointy-haired bosses to FOSS

Mike Dailey tries to convince us gave me the impression that moving to Linux is an all-or-nothing proposal. That's plain ridiculous. An undertaking of that scale is a gigantic project, which no CIO will endorse. What you can do is introduce FOSS technology in the enterprise, step by step. Once it has proven itself you may take the next step. Since mixing Microsoft and FOSS is a viable scenario you have a multitude of options at your disposal.

Warranty and support
The plain truth is that although most managers may have heard of Open Source, they don't have a clue what it is or how it works. Most of them think it is something like public domain software, stuff you can get for free, without warranty and without support. So the first thing you have to do is to tell them you can get support from reputable parties like Oracle, IBM, Novell and RedHat. If you're not happy with their support, you can change with more ease than you could with closed source software.

Some CIOs are completely unaware that these parties provide updates. They think you have to monitor a multitude of websites or CVSes, search for the newest versions, download a tarball and recompile it. I always tell them jokingly: "Well, I don't know what I've been getting all these months, but it seemed like patches".

Note that very few CIOs know that Microsofts warranty is rather limited. In fact, it is limited to the smallest extent that law will allow.

Code quality
Most CIOs think that Open Source is produced by amateurs, hacking away in their attics and garages, so they have concerns about the code quality. Of course, you can tell them that reputable companies produce a vast amount of FOSS as well, but that doesn't make them any happier. May be this will help. Okay Mike, I got the message: anybody can come up with some stray links. So especially for your reading pleasure I have delved a little deeper into this subject.

According to Carnegie Mellon University's CyLab Sustainable Computing Consortium closed source software contains 20 to 30 defects per thousand lines of code (KLOC). According to Steve McConnell, the best a software company could achieve is 0.1 defect/KLOC. That is, if you apply the highest standards known to man to the software engineering process, which is equivalent to CMMI level 5. And how many companies have achieved this walhalla of software engineering? 21. Yes, you're reading it correctly: 21. Over 99% of the software companies in the world are still in what SEI calls "the anarchy and folklore" stage.

But let's focus on this 0.1 defect/KLOC figure. It is very hard to find how well Microsoft is doing. I've spent hours trying to find any figures and I finally found them: it's between 0.5 and 1.8 defects/KLOC, depending on the methods used.

During a training at SEI, Microsoft engineers managed to get their defects rate down from (hold on to your hats) 25 defects/KLOC to 7 defects/KLOC. At the end of the training, they brought their defects rate down to an incredible 0.06 defects/KLOC. But that was during a training. Not a death march.

But for arguments sake, let's hold on to this incredibly low 0.06 defects/KLOC figure, which even baffles the most sophisticated software companies in the world and turn to FOSS. Coverity, a company specialized in software integrity products, has been evaluating FOSS projects for several years. Here you can see their figures. All Rung 2 projects surpass this 0.06 defects/KLOC standard easily and about 40% of the Rung 1 projects match or surpass the 0.06 defects/KLOC mark. The highest defect rate still matches a CMMI level 3, a standard which 99% of the closed source software companies have not been able to achieve.

So yes, I'm fairly confident about the quality of FOSS software.

Vendor lock-in
Closed source companies have a commercial interest in limiting your choices. They will only allow interoperability if it doesn't affect their sales. Consequently, applying their solutions often results in silos, repositories that function very well in their own right, but don't communicate very well with the outside world.

Mike Dailey can't know this, because it is not his line of expertise, but the information architecture of most companies is abysmal. Too much information is locked up in documents and spreadsheets, information that would better be served by being stored in repositories. The truth is that it requires expertise and a vision that most CIOs simply do not have. It requires an Enterprise Architecture, something I have rarely found in the companies I worked for.

Since the dependence of office suites is so great, replacing them is no walk in the park. Entire applications are built on top of office suites, applications that enterprises are depending on. This gives you very little leverage when negotiating your next deal with Microsoft. You're stuck. And the more software you buy to relieve your problems, the harder it gets to turn around and get away.

Another disadvantage may not be so evident, but I've seen it happen. Some of these applications depend on a certain version of Microsoft Office. They were written by some employee at some moment in time and badly documented, so modifying it is not a viable option. However, if you want to have the next version of Sharepoint you need the newest version of that very same office suite. It's a nightmare: you cannot upgrade and you cannot downgrade. What are you going to do?

Even in a scenario like this FOSS offers a solution. You can always compile an application for your current platform and run it along with your new version. Sure, you may not have any support for it, but your business will continue. The same applies if a FOSS company goes belly-up. With closed source software you need complex escrow procedures to accomplish the same feat.

Finally, FOSS is open by definition. If you want to achieve interoperability, you can do just that, since there is nothing going to stop you. Most closed source software prohibits you to even read the repository. No more silos!

Scalability
When you choose Windows, you have the choice of Intel, AMD and.. that's it! When you choose Linux you have a clear upgrade and downgrade path, from the tiniest netbook and cellphone up to the mightiest mainframe and supercomputer. Just name any platform, Linux runs on it.

Did your boss know that 85% of the worlds supercomputers run on Linux? Did he know that she runs on Linux? Did he know his TomTom (still) runs on Linux? Did he know even his Android cellphone is in fact running Linux?

Licensing
Steve Ballmer has done a good job. Microsoft has always been good when it comes to marketing and spreading FUD. Most managers are afraid to use FOSS, because they fear their custom applications will automatically have to be released under an Open Source license. Of course, this is not true, but you may have to deal with this issue.

It hasn't been done
Most CIOs are unaware that they already use FOSS. The IBM HTTP server for instance is a direct descendant of Apache. There are various successful projects in Europe you can refer to. Recently, French police switched from Windows to Linux. Several German cities have switched from Windows to Linux.

It's just as expensive
Even if that were true, there are various advantages to FOSS which make a perfect business case. Like better interoperability, scalability, no vendor lock-in, which will make FOSS an attractive option. And next time your local Microsoft representative comes along, your position in the negotiations has significantly improved.

Most companies pay too much for their closed source applications, because their Configuration Management is not up to par and collecting data on actual installations is expensive and cumbersome. FOSS licensing is usually much more transparent. And you don't have to be afraid of the BSA anymore.

FOSS in a Windows world
Don't try to transform your Microsoft shop all at once. You won't succeed. A new project utilizing PHP (which is also supported by Microsoft) may be a good start. Some SQL-Server sites can easily be replaced with MySQL. Replacing IIS with Apache and Windows with Linux becomes much easier after that.

Another scenario is replacing Microsoft Office with StarOffice or OpenOffice, without resorting to a completely FOSS workstation. Visio and MS-Project are notoriously expensive applications, but Dia and OpenProj will do just as well. You can easily exchange data by using SVG or XML formats. Again, when most closed source applications have been replaced with FOSS equivalents, moving to a full fledged FOSS platform becomes much easier.

There may remain pockets of closed source and frankly, I don't think you will be able to remove those very easily. The point is, do you really need to. SAP for instance, is an application that runs very well in a FOSS environment, and so does Oracle. The point is you have given the enterprise more choices and tipped the balance in favor of FOSS.

So what do you think, Mike, is the Linux debate really dead? I don't think so. Thank you for proving this.

Update: The rumours of the death of the debate have been greatly exaggerated: a new article has been published by John Buswell, giving an entirely new view on the subject.

Update: Mike Dailey has truly given a worthy closing to this debate. Although I cannot undo my previous post, I cannot honestly maintain that Mike Dailey fits this profile.