Linux.com

Everything Linux and Open Source

The true cost of software ownership

January 17, 2004 (8:00:00 AM)  -  5 years, 10 months ago

By: David Sugar

What does "total cost of ownership" really mean in relation to commercial computer software? It's not the same as ownership of a tangible asset. I own a '67 Chevy -- I have a title that says so, and all of the rights and responsibilities of ownership that naturally comes with it. I can compute what I paid to acquire and maintain this asset in usable condition.

There is no ownership involved in proprietary commercial software, because all users are mere licensees. Users are at best tenants, and the most correct and appropriate term should be "total cost of usage" (TCU), since commercial proprietary software is now often sold on leases.

When we talk about commercial free or open source software and try to determine the cost associated with its use or whether there is a TCO, we find that we have something fundamentally different from proprietary software. Free software often comes with a license, like the GNU General Public License, or others, that grants rights that feel a lot more like ownership. This is not to say that "ownership" as such actually exists in software, whether free or proprietary; this is not supported in fact or in law due to the intangible nature of software. But setting aside the question of whether software has owners in the conventional property sense, users of commercial free and open source software do receive an owner-like interest in the software that they are free to exercise, and so the term "total cost of ownership" (TCO) may potentially be applied.

To understand why software cannot be owned, except through artificial agreements and contracts that restrict software to make it conform to unnatural constraints, one need only consider the nature of intangible goods. Intangible property is non-rivalrous. While I can look out the window and see that I have that '67 Chevy on cinder blocks rusting on my front lawn, I can be sure that nobody else does. If somebody else did choose to take it overnight while I was asleep, bless their soul, it would be gone, and I would be denied the ability to use it.

In the case of software or ideas, the fact that a program or idea exists in my home does not prevent the very same program or idea from existing or being used simultaneously elsewhere. Even if the software or idea is borrowed, it does not mean that it vanishes from my machine or mind, never to be used by me again.

In place of property ownership we have the concept of copyright for these intangible works, which puts some temporary artificial constraints and offers specific privileges to an "original author" when expressed in a specific form. Copyrights are designed to promote the long-term public good of having more original works created and published by offering temporary monopolies in their use as an incentive to publish.

Richard Stallman likes to say "software has no owners" for a number of reasons. However, what free software licenses like the GNU General Public License do is permit many of the privileges of copyright to be shared with everyone. This is somewhat like the idea of sharing ownership, and is certainly far more natural, given the nature of software, than some of the artificial constraints found through copyright law today. Hence, this shared privilege of copyright we can substitute as comparable to ownership and hence have basis to say TCO may indeed exist for Free Software.

How do we compare the TCO of Free Software with the TCU of proprietary software? Well, we can look at costs, such as license fees and training and support costs. Proprietary software vendors often claim that license costs represent only 20% of the total cost of usage when considering these other factors. However, as we will shortly see, TCU for proprietary software is very different from TCO in free software.

The first problem one finds with TCU is that it is a continually re-occurring cost. Microsoft loves to claim that studies show that the TCU of its proprietary software, often over a period of three years, is not much different than the TCO of free software over the same period. However, there is a reason they often choose a three-year period for their studies; they license the use of their software under enterprise agreements for three-year periods, and in the fourth year, users have to pay for additional time. With Free Software, there is no re-occurring cost to license or use the software.

However, the TCU for Microsoft products are even worse than this over the long term. Microsoft's enterprise agreements require and force licensees to purchase and use the latest release of its software, and restrict the ability for licensees to acquire old releases -- a policy known as "forced upgrades." This means that every few years, their customers' users and support staff have to be retrained to use new and different products at the vendor's whim. This means that customers not only have to perpetually purchase new software licenses, but also all of the training and support, over and over again.

Customers may also find that with new software, existing custom and in-house software may no longer function or perform as well as before, since old APIs are being emulated, hence this in-house software may also need to be rewritten or at least updated, adding additional cost. Hence, the annual TCU of proprietary software under renewable enterprise agreements never significantly diminishes over time.

By contrast, with Free Software, nobody forces someone to change the software they have deployed. One can choose to upgrade free software, in part or in whole, at any time they desire, or choose never to do so. When working systems are left in place, the cost of supporting them goes down over time, hence the annual TCO of Free Software actually decreases over time.

When I was involved in federal contracting, we often computed TCO for contracts over a period, not of three years as in many of Microsoft's published case studies, but rather over 10 years, which was a more realistic perspective for true enterprise costs over time. Over a 10-year period, what we often found was that the TCO of Free Software, even computed through standard means, was about 1/2 to 1/3rd the TCU of comparable proprietary software. That was before we looked at all the additional liabilities that proprietary software burdens the enterprise customer with, such as losing the right to resale, the possible unexpected costs of license audits, and many other risks too numerous to enumerate here.

Free Software TCO is a lot more like true ownership, where once you purchase something, you do not have to purchase it again and again, and once it is payed off, it too can serve you very well for many years -- until it is ready for the cinder blocks.

Read in the original layout at: http://www.linux.com/archive/articles/33760