Ask a parametric designer what a building is and you get an honest answer that surprises most clients: it is not a model. It is a federation of models — the architect’s in one authoring tool, the engineer’s in a second, the MEP consultant’s in a third, the facility manager’s in a fourth — and the real object of the discipline is the seam between them. Interoperability is the study of that seam. It is not a feature you switch on; it is a geometric and semantic problem you solve, again, every time a model leaves the building.

←TODAY: In 2026 two IFC exports of the same steel frame can both be 100% ISO 16739-compliant and still describe two different buildings. →3012: By the Zurich-3012 horizon the geometry travels losslessly and the derivation travels with it — the seam carries the why, not just the vertices. Fulcrum: An open format guarantees the bytes are readable; it never guaranteed the meaning survived — and meaning is the only thing worth keeping.

What it is: BIM interoperability is the disciplined transfer of a building’s data between tools that were never designed to agree. The carrier is buildingSMART’s Industry Foundation Classes (IFC) — a neutral, vendor-independent schema that became ISO 16739 in 2013 — paired with ISO 19650, the process standard (published December 2018, out of UK information-management practice from around 2007) that governs how you pass the data around. One standard for the thing, one for the way you move it. Together the industry calls them openBIM: not a product you buy, but a promise that the model you author today still opens in thirty years on software that does not yet exist. PAZ’s own concept library puts it precisely — interoperability is “a methodology, not a monument.” Its landmarks are documented hand-offs, not signature facades.

Why it works: The mechanism is easiest to see if you think like a form-finder. A building model is a graph — nodes (elements) carrying attributes (profile, material, placement), edges (relationships) carrying topology (this beam bears on that column, this space is bounded by those walls). IFC works because it defines that graph in a schema every tool can read: an IfcBeam is an IfcBeam whether Revit or Archicad wrote it. What IFC cannot enforce is the encoding choice. The same concrete column can travel as a swept solid, a boundary representation, or a tessellated mesh — all three legal, all three read differently by the receiver. The peer-reviewed assessment of monodimensional-member structures in Applied Sciences (2021) documented exactly this: profile identities, materials, and placements that arrive degraded or dropped between tools that both claim compliance. This is why the profession lives at BIM Level 2 — a federation of verified exchanges — rather than the single shared database of Level 3 / iBIM. The structural logic to internalise, as PAZ’s engineering concept panel states it flatly: verification is not overhead on top of interoperability — verification is interoperability. The geometry travels; the assumptions — load combinations, mesh settings, boundary conditions — do not, and must be re-checked by hand after every round-trip.

Origins: BIM was born not as software but as an argument that ran quietly from the 1970s: describe a building once, as data, and draw it many times. The term settled only in the early 2000s. The turning point has a name — the Industry Foundation Classes, first shipped as IFC 1.0 in June 1996 by the body now called buildingSMART International (founded 1994 as the IAI). IFC2x3 (2006) was the first version most of the industry actually exchanged; IFC4 modernised the geometry and semantics; IFC4.3 / ISO 16739-1:2024 finally stretched the schema out of the building and into infrastructure — rail, roads, bridges. The political turn came when governments made the open model load-bearing for public money: Denmark mandated IFC for publicly aided projects in 2010, Finland’s Senate Properties followed in 2017, Norway’s Statsbygg across all projects. A schema became a procurement condition — proof that the open format was an argument about ownership, that the data of a building belongs to the building over its whole lifecycle, not to the vendor who happened to draw it.

In practice: Where does a Swiss studio actually reach for this? The clearest reference is the EU Horizon RecycleBIM consortium (2021–2023), which authored interoperable IFC models across both Autodesk Revit and Graphisoft Archicad to carry material data through a circular-construction workflow — the whole point was that one building had to stay legible to two rival tools and a downstream material-management process. Closer to home, EPFL’s open-source BIM initiative in Lausanne treats the toolchain itself as a public good rather than a licence, which is exactly the ground on which interoperability is won or lost. For a working office the leverage point is Rhino and Grasshopper as the connective tissue: David Rutten’s Grasshopper (2007) lets you read, remap, and re-emit model data between ecosystems as an explicit, auditable graph rather than a black-box export button — and PAZ Academy’s own Grasshopper↔Archicad interoperability work exists precisely to fill the missing parts of that seam for DACH practices. The habit to build is hygiene: every model leaves as IFC first and native second; every model that arrives is opened with a checker before it is trusted. And be candid about the fact that Level 2 is not a compromise waiting to be upgraded — it is, for now, the professionally honest way to work.

Hack: Diff two IFC exports member-by-member and watch which profiles survive the round-trip. Export the same tiny frame — two columns, one beam — from two authoring tools, then let ifcopenshell (Thomas Krijnen’s open-source IFC toolkit) read both and print the disagreements. The diff is the interoperability gap, made visible.

import ifcopenshell
def beams(p): return {m.Name: getattr(m, "ObjectType", None) for m in ifcopenshell.open(p).by_type("IfcBeam")}
a, b = beams("frameA.ifc"), beams("frameB.ifc")
for k in sorted(set(a) | set(b)):
    print(("OK  " if a.get(k) == b.get(k) else "DIFF"), k, a.get(k), "->", b.get(k))

Run it once and “open format” stops being a slogan. You will watch a profile name change or a member vanish between two tools that both claim IFC compliance — and you will never again confuse open with lossless.

There is a longer shadow here worth naming plainly. The parametric work my generation came to regret was rarely the ugly form — it was the model whose logic nobody could reconstruct after the plugin went dark. A geometry with no derivation is a beautiful guess you cannot defend in a structural review, and a federated model held together by an undocumented export ritual is the same failure at building scale. The buildings drawn now will outlive three generations of software. The one guarantee that they stay readable is an open format policed by people who actually check the seam. So keep the standard written down: which schema, which MVD, which property sets carried, which assumptions were re-verified after each hand-off. Keep the maths; the file format will not survive you.

The frontier is not a bigger single model — it is a thinner, more trustworthy seam. As the vendor bridges mature and IFC’s structural coverage tightens, the manual re-verification will shrink, but it will not vanish, because meaning is harder to standardise than geometry. Own the standard, automate the diff, and treat every hand-off as a question you answer with evidence. That is how a building becomes independent of the tool that drew it — and that independence, not any single feature, is the whole point of BIM.