1 — BETTING ON THE WEB
JLP Décryptage
When you founded Globalis in 1997, you chose to specialise in web development and PHP at a time when that ecosystem was still taking shape. What convinced you that this technology and the Web’s open philosophy could provide lasting foundations for a company?
Frédéric Hovart
In 1996, the Web was still in its infancy, but my future business partner and I were convinced that it was going to profoundly change the world.
It opened up a space for the circulation of knowledge and made it possible to create services accessible from any computer. A small company could now address a global market, and organisations could connect directly with their customers. We were convinced that this market was going to explode and that there were opportunities to be seized. At the time, few players had made that bet, particularly in France.
I then left a well-paid job abroad, in oil exploration, to create Globalis.
The choice of PHP and MySQL emerged gradually between 1997 and 1999. These technologies enabled faster development, were quick to learn and met the needs of dynamic web applications well. Their open philosophy also appealed to us in contrast with the closed environments of major software vendors. However, we never regarded them as an end in themselves. Globalis’s purpose is to help organisations make the most of the Web. PHP was a means, and a means must be able to evolve. PHP and MySQL were tremendous growth drivers for Globalis, but today account for only a minority of our revenue.
2 — BUILDING A TECHNICAL COMMUNITY
JLP Décryptage
Globalis was involved in the birth of the AFUP and the first PHP Forums, at a time when this technology still had to establish its legitimacy among businesses. What does a technical community bring to a language that companies, schools and vendors cannot produce on their own?
Frédéric Hovart
A community transforms a language into a genuine ecosystem. It produces the documentation, libraries and tools that make it usable on a large scale.
Within a community, events bring members together and popularise the technology. They give the language legitimacy among businesses by showing that it does not depend on a handful of isolated developers.
In November 1998, we created PHPindex.com to inform, unite and support the French-speaking PHP community. The site became our number-one salesperson and generated a significant stream of business for nearly 15 years.
In the early 2000s, together with other pioneers, we organised the first PHP Forums and then founded the AFUP. These two entities contributed greatly to the growth of PHP in France. They still exist today and rank among the largest organisations in the French-speaking open-source world.
3 — KNOWING WHEN TO LET GO OF WHAT YOU CREATED
JLP Décryptage
In 1999, Globalis developed Carbone, a framework designed around very tight memory and processing-time constraints, which would support its projects for around fifteen years. How can a company recognise that an internal tool—even one that has proven itself and is closely bound up with its history—must now make way for other architectures?
Frédéric Hovart
It is important to remember that in the early 2000s, PHP frameworks did not exist. We had understood the value of these reusable libraries: they enabled us to speed up the development of our projects and make it more reliable. Carbone stayed with us for around fifteen years and was a great success.
Frameworks gradually began to appear. Some offered more modern architectures, more features and much larger communities, giving them formidable momentum. The hardest part then is not to confuse attachment to what you have built with its actual usefulness. Abandoning Carbone did not mean that we had been wrong to create it. It had simply fulfilled its purpose.
The transition was gradual: we maintained existing applications on Carbone—some still exist today—and adopted other architectures for new projects. A technology company must know how to build on its history without becoming its prisoner.
4 — GROWING WITHOUT BECOMING GENERIC
JLP Décryptage
For nearly thirty years, Globalis has operated somewhere between a web agency, a development shop and a digital services firm. How have you preserved a strong technical identity while meeting the growth, organisational and profitability requirements of a fifty-person company?
Frédéric Hovart
The situation varies depending on the service.
In technical assistance, growth is relatively straightforward to manage: it depends primarily on our ability to recruit the right skills and maintain a satisfactory staffing level.
The situation is more delicate for consulting, development and maintenance of websites and web applications. Growing while maintaining a high level of quality means training new hires and structuring teams, methods and project management without losing the technical standards that define our identity.
Our growth was steady and gradual for twenty-five years. Globalis remained a digital services firm specialising in web and mobile development, with strong technical expertise and constant attention to the quality of its deliverables. We preferred genuine mastery of a few areas to a very broad catalogue of services. We also sometimes turn down business when we are not certain that we can achieve the client’s objectives under the right conditions.
For a long time, the economic equation for a digital services firm was fairly simple: satisfactory staffing, daily rates consistent with salaries and well-managed projects ensured profitability.
That balance is far more fragile today.
The accumulation of geopolitical and institutional crises is curbing investment by clients in our sector.
Competitive pressure from low-cost countries is increasing.
AI is completely transforming the methods and speed of software production.
These obstacles can sometimes become opportunities that must be seized. To do so, our profession must reinvent itself in depth. The subject is absolutely fascinating, and a new Globalis will emerge in the coming months.
I also know that many leaders of digital services firms and digital agencies are asking themselves the same questions. I already organise lunches and peer experience-sharing sessions, and I am naturally available to talk with those who wish to test their thinking against my experience.
5 — PASSING ON TECHNICAL RIGOUR
JLP Décryptage
You regularly emphasise training, code review and supporting developers when they join. In a sector where tools evolve very quickly, how do you pass on not only skills, but also a lasting commitment to code quality and maintainability?
Frédéric Hovart
Things have changed considerably since 1999, both in recruitment and in the training of technical teams. We still devote just as much time to recruitment. We look for developers who understand what they are doing, can explain their choices and are genuinely committed to their work.
For the past few years, every new hire has followed a structured onboarding programme, with regular check-ins to assess their progress. Mentors support them. Pair workshops are organised to facilitate the transfer of Globalis’s practices and culture. Code reviews are systematic.
We state clearly that the code and the project are a collective responsibility: everyone must be able to understand, take over and develop the work of others. We also ask teams to step back from what they produce, beyond mere technical functioning. Keep it simple; avoid overengineered systems that cannot be maintained…
We then offer a whole range of resources from which teams can choose to progress: book purchases, conferences, knowledge sharing, time dedicated to technology monitoring or training... We also provide a time allowance for those who wish to teach at university. The circle is thus complete.
6 — KNOWING WHEN TO SAY NO TO AI
JLP Décryptage
During your webinar on artificial intelligence applied to web projects, you focused on concrete use cases rather than the broad promises surrounding this technology. What criteria allow you to determine that a project genuinely calls for AI, rather than a more conventional, simpler and sometimes more reliable solution?
Frédéric Hovart
Yes, the webinar addressed the integration of AI features into web projects. Sometimes AI is a good idea; sometimes it is not.
We always start from the business problem, never from a desire to add artificial intelligence to a project.
The intense media attention surrounding AI should not influence the technical decision. We are not here to indulge ourselves, but to provide the most relevant answer to the client’s need. Artificial intelligence is a revolutionary technology that opens up considerable possibilities. It is obviously not the solution to every problem, and recognising that is simply common sense. AI is a means to an end; it is not an end in itself.
Everything then depends on the problem at hand, the available data and the expected level of confidentiality, the required level of reliability and the consequences of a potential error.
If a conventional technology better meets the need, we must know how to choose it and say no to AI.
7 — BUILDING IN REVERSIBILITY
JLP Décryptage
Globalis says that it transfers ownership of developments to its clients and avoids locking them into contractual or technological dependence. How does a service provider build a lasting relationship while giving its client precisely the means to one day do without it?
Frédéric Hovart
Globalis’s clients are free to change service providers at any time. In practice, this remains rare, since more than 95% of them continue to work with us year after year.
This loyalty rests on the quality of the work and on trust, not on engineered dependency. We chose open-source technologies and apply the same philosophy to our contractual relationships.
When a client decides to leave, we hand over the source code, databases and all the necessary documentation. Everything belongs to the client. We also undertake to support the new team through a genuine transition service. This handover is conducted professionally and enables the project to continue under the right conditions. It is not always easy to experience, but that is our mission.
For us, the best way to build a lasting relationship is to give the client the freedom to stay or leave. We have always been very proud of the length of our engagements; more than 10 years is quite common, and that is the greatest reward.
8 — WHAT MUST NOT DISAPPEAR
JLP Décryptage
If you could pass on just one thing to developers who have only ever known the Web with frameworks, the cloud and now generative artificial intelligence—a skill, a mindset or a form of vigilance—which would you choose?
Frédéric Hovart
I would first pass on the requirement to understand what they are doing. Tools simplify many tasks, but can also conceal a complexity that we must remain able to analyse.
We must retain perspective, question the answers proposed and preserve our own judgement.
Curiosity and the ability to continue learning are absolutely fundamental. Our profession will change in record time, and only those who know how to evolve will be able to take full advantage of this transformation.
