Your WinDev application works. It carries ten or fifteen years of business rules that nobody ever wrote down anywhere but in WLanguage, and data your business cannot afford to lose. Migrating to .NET is not rewriting a piece of software: it is moving those two things, intact, to a platform you will be able to keep alive. Here is the method I apply, as it served on a field sales CRM deployed on about 110 workstations.
Why leave WinDev
The reasons are the same from one client to the next, and they are rarely technical at first.
- Dependence on a single vendor. The environment, the language, the HFSQL database and the licensing model all depend on one company. Every commercial decision of the vendor becomes a decision imposed on you.
- Recruiting. Finding an available WinDev developer has become hard; training one on a fifteen-year-old codebase is harder still.
- Licence costs, which weigh all the more when the application runs on many workstations.
- The need for web and mobile. Your customers, branches or technicians want to reach the application from a browser or a tablet, and the original architecture was never designed for that.
None of these reasons requires throwing everything away. They require a method.
Step 1: map the real code and the real schema
The source is the truth. Before writing a single line of C#, I inventory what actually exists: windows and procedures, classes, queries, scheduled jobs, and above all the HFSQL analysis as it is in production, not as the documentation describes it. No field, no table, no rule is invented or assumed: everything that goes into the target was read in the existing system.
This is where you find business rules buried in a button’s code, fields diverted from their original purpose and tables nobody reads anymore. Finding them later, during acceptance testing, costs ten times more.
Step 2: scope with traceable architecture decisions
Every structuring choice is written down in a short, dated architecture decision record (ADR): WPF or web target, SQL Server or PostgreSQL, authentication through the company directory, excluded scope, equivalence criteria. These notes prevent re-debating a settled point three months later, and they become the documentation the next team will read.
On the field CRM, scoping set a .NET 10 WPF target with an MVVM architecture, LDAP / Active Directory authentication to remove application accounts, and direct integration with the Infor M3 ERP for items, customers and orders.
Step 3: rebuild incrementally, testing each step
A single big-bang switch is the surest way to lose six months and the users’ trust. I proceed in batches, and each batch is tested before the next.
- Data first. A .NET importer reads HFSQL and fills a staging area faithful to the source, then a modernised schema. Every table is checked.
- Headless processes next: extractions, synchronisations, exports. They are easy to compare side by side with the old program.
- Screens last, module by module. The WinDev application stays in service until each scope is switched over.
Why not a lift & shift, an identical copy? Because it would carry over the limits of the old model. Import faithfully, then modernise the schema: correct types, keys, constraints. Fidelity is a step, not a goal.
Step 4: harden
The .NET version must be safer than the old one, not just newer. Three rules without exception:
- No secret in clear text. On an Infor CloudSuite invoice extraction tool rewritten in .NET 8, passwords are encrypted with DPAPI and no longer appear in any configuration file.
- External access in read-only mode wherever the application does not need to write.
- Code signing, so that workstations only run what you delivered.
The same extraction tool illustrates another gain: its multi-process orchestration, inherited from the WinDev program, was replaced by asynchronous processing, easier to monitor and to resume after an incident.
Step 5: industrialise deployment
This is the step projects forget, and the one that makes the difference on a fleet of a hundred workstations. An MSI installer, automatic updates at launch, a signed delivery. On the field CRM, deployment on about 110 workstations required no intervention on any individual machine: the sales rep opens the application, it updates itself, they work.
Step 6: maintain and hand over
Up-to-date documentation, recorded architecture decisions, training of the internal team or of the provider who takes over. A successful migration project is one you can walk away from without anything stopping.
Why the incremental approach keeps users on board
A migration is judged by the people who use the application every day, not by the architecture diagram. Working in batches gives them something concrete to validate every few weeks: a data set they can query, an extraction they can compare with yesterday’s, a screen they can use on real cases. Their remarks come back while the context is fresh, and they shape the next batch instead of piling up for a final acceptance phase nobody has time for.
The coexistence period has a cost, since two applications and two databases must be kept in step, but it is a cost you control. You decide when each scope switches over, and you can pause between two batches if the business calendar requires it. A single big-bang switch offers none of that.
What you can do right now
If your WinDev application is at the heart of your business, the first decision is not “migrate or not”, it is “know exactly what we have”. Mapping and scoping fit in a short remote audit that gives you a roadmap and comparable scenarios. You will find the details on the services page and, if the PC SOFT runtime fee is on your mind, the possible options in PC SOFT runtime fee: what options for WinDev software vendors?.
To go further: the WinDev migration page gathers the six-step method, the figures from completed projects and a man-day estimation simulator.