Free software is entering Italian public-administration tenders as a technical option to be assessed, no longer as an academic curiosity. The commission the Minister for Innovation and Technology set up in January 2003, chaired by Angelo Raffaele Meo of the Politecnico di Torino, has exactly that brief: to work out whether and how open source code can enter public information systems. I write as someone who builds solutions with free software for his own clients, after a day spent discussing the matter in Pisa.

Context

The debate ran aground at once on the cost comparison: proprietary licences against free distributions. That is slippery ground. The licence price is a visible line item, handy to put in a spreadsheet, but it almost never weighs more than the rest in the total cost of an information system: data migration, training, maintenance, and integration with what is already running are worth far more, and have little to do with the licensing model.

Chasing the licence price draws attention away from the two things that genuinely weigh on an administration: control over the data formats and control over the code that processes them. These are what decide whether, ten years on, those records will still be readable, and whether that service will still be maintainable by someone other than the original sole supplier.

The lock-in problem

An administration that stores acts, registries, and case files in a closed, undocumented format leaves to a single supplier the decision over when, and at what price, those records remain accessible. The constraint is not contractual but technical: even once the contract has lapsed, the data stays hostage to the program that wrote it.

Free software does not untie this knot merely by being free. It unties it when it carries open, documented formats and when the source is in fact available to a second supplier. These are two distinct requirements. There is proprietary software that saves in standard formats, and free software that uses poorly documented formats of its own. The licence is needed but not enough: the concrete guarantees are the format specification and the legibility of the source.

For public administration this distinction weighs in a particular way. A private party tied to a supplier suffers a contained economic loss. An administration that loses access to its own archives compromises the continuity of a service owed to citizens, and does so with public money.

What the open-source model offers

The established free licences set down precise guarantees in black and white. The GNU General Public License, in its version 2 of June 1991, obliges anyone distributing a program or a derivative of it to make the source available under the same terms. The BSD family is more permissive and allows redistribution in closed form as well. The Open Source Definition, kept by the Open Source Initiative since 1998, sets the minimum criteria a licence must meet to qualify: free redistribution, available source, permitted derived works, no discrimination as to field of use.

For an administration the practical consequence is being able to put maintenance out to tender. If an application’s source sits under a free licence, the body can assign fixes and further development to a supplier other than the one who wrote it, without starting over. It no longer depends on a single party, but on the technical competence available on the market.

Here a space opens for small and medium Italian firms. An SME hardly holds up against a large supplier on proprietary licences, where scale wins. It holds up on the maintenance and adaptation of free software, where knowledge of the application domain and proximity to the client win. It is a market of services, not products, and it is played out on competence more than on volume. This very perspective of SMEs building professional solutions on free software is the one noze brought to the conference with the talk “Open Source at work: the offering of Italian SMEs” (https://www.noze.it/en/insights/oss-pa-conference-pisa/).

Critical point

The real risk is that the openness stays on paper. A body may adopt an open-source application and still find itself tied to a single supplier, if that source is written so as to be illegible to anyone else: no documentation, undeclared dependencies, a build that does not reproduce, configuration scattered and implicit. In that case the licence is free on paper, yet the technical lock-in stays whole.

Having the source is a necessary condition, not a guarantee. For it to bear the expected fruit, verifiable practices are needed: documentation of the data format, a reproducible build procedure, deposit of the code with the administration and not only with the supplier. The circulars issued from 2001 by the then AIPA on the reuse of public applications went this way: they required that software developed with public money be reusable by other bodies. And for it to be so, before any licence, the code has to be put in a condition to be reused.

Implications

Across Europe a few experiments are testing these hypotheses in the field. Since 2002 the Spanish region of Extremadura has distributed in its schools tens of thousands of workstations running a purpose-built GNU/Linux distribution, gnuLinEx. The city of Munich is at present running a study to choose the successor to its Windows NT 4 infrastructure, and the assessment weighs, explicitly, a proprietary path against a free-software one. These are cases to watch for the data they will produce, not to cite as settled proof: the real figures on migration and maintenance will come over the years, not in the press releases.

On method, one useful point has already emerged. In 2002 the Peruvian congressman Edgar Villanueva, replying to a Microsoft letter on his bill for free software in public administration, shifted the matter onto principles: for an administration, availability of the source, access to the data, and independence from the supplier are requirements of good government, and they hold regardless of any price comparison. That framing holds up better than the licence-cost arithmetic, because it does not hinge on how a price list swings.

Limits

Not everything fits. Highly vertical software, with few users and no development community behind it, will hardly find in the open model the distributed-maintenance benefits that apply to base components — operating system, servers, office suite, databases. There the advantage is clear and daily use already confirms it. For specific line-of-business applications the choice has to be weighed case by case, on the individual tender.

And the question of in-house competence remains open. An administration that adopts free software without staff able to read a technical specification, judge the quality of a source tree, and demand documentation and build procedures, merely shifts its dependency from one supplier to another. The lever is not the licence: it is the body’s capacity to exercise the control the licence puts in its hands.

Cover image: Stylized illustration of a bearded, horned gnu head, the GNU Project logo — logo by Aurelio A. Heckert, CC BY-SA 2.0 — https://commons.wikimedia.org/wiki/File:Heckert_GNU_white.svg