Design Software History: Visualizing Engineering Intent: Feature Histories, Constraints, and Semantic PMI in CAD

December 29, 2025 16 min read

Design Software History: Visualizing Engineering Intent: Feature Histories, Constraints, and Semantic PMI in CAD

NOVEDGE Blog Graphics

What “engineering intent” looks like on screen

Beyond geometry: features, history, and semantic control

On a modern CAD canvas, the most revealing artifact is not the shaded model but the scaffolding that produced it. Parametric systems expose this scaffolding through a feature tree that encodes sketches, profiles, and procedural operations in a time-ordered history. Each sketch declares its primacy with driving dimensions, constraints, and references; each extrude, revolve, sweep, or loft documents how shape arose from rules. The hierarchy matters: reordering a fillet beneath a shell can invert manufacturability, and suppressing a rib can trigger regeneration failures that trace dependency lines back to their sketch parents. Visualization helps here: dependency graphs, often drawn as node-and-edge maps, highlight parent-child relations and circular references. Even selection is intentful—topology-aware hit-testing lets designers pick edges as feature results rather than mere tessellation, so a chamfer can be edited regardless of mesh density.

These cues are reinforced by glyphs and overlays. Sketch planes are tinted, profiles thicken when closed, and driving dimensions glow when active. Systems from PTC Creo (heir to Pro/ENGINEER), Dassault Systèmes SolidWorks, Siemens NX/Solid Edge, and Autodesk Inventor popularized dynamic dimension editing at the cursor, where dragging an arrow updates both the value and a preview of downstream features. In direct modeling contexts, the “feature” may be inferred retroactively—Siemens’ Synchronous Technology and Autodesk Fusion 360 detect patterns, holes, and symmetry, visualizing editable procedural intent even on dumb solids. Across these tools, the insightful elements are the ones that expose not just what exists but why it exists, collapsing the gap between model and method through visual feedback such as timeline markers, regeneration warnings, and feature-scope highlights that illuminate participating faces.

  • History bars with roll-back handles visualize regeneration order and afford quick “what-if” trials.
  • Sketch entities render with distinct styles (construction lines vs. geometry) to separate scaffolding from shape.
  • Dependency graphs and parent/child callouts spot unintended couplings before they metastasize.

Constraints made visible: degrees of freedom, mates, and status color

Constraints are the language of intent, and their on-screen epigraphy makes models legible. In sketches, horizontal, vertical, perpendicular, tangent, and coincident constraints draw as compact glyphs at entity junctions; applying them collapses the model’s degrees of freedom, a change made tangible by DOF triads and arrows that disappear as mobility vanishes. Many systems use color to signal constraint status: under-constrained profiles show in one hue, fully constrained in another, and over-constrained (or conflicting) situations pulse or cross-hatch. This color vocabulary—popularized by widely licensed solvers such as D-Cubed’s 2D DCM and 3D DCM (authored under John Owen and now a Siemens asset)—became a de facto standard across SolidWorks, Autodesk, and Unigraphics/NX, giving users an immediately recognizable grammar for geometry health.

Assemblies speak a related dialect through mate glyphs: planar mates, concentric constraints, gear relations, and path followers display as icons attached to components, often with drag-arrows that visualize residual motion. When a hinge is properly constrained, only a single rotational DOF arrow remains; when an axial slider still floats, a linear arrow invites a designer to add a lock or limit. Constraint graphs, sometimes presented as collapsible trees or network diagrams, reveal how parts interlock and which subassemblies define reference frames. In editing mode, mating faces highlight with transparent colors and edges thicken, avoiding ambiguity in densely packed contexts. The model becomes a machine-readable sentence composed of verbs (mates) and nouns (features), and the UI’s emphasis on glyphs and colors is what teaches that sentence to the eye.

  • Mate glyphs and residual DOF arrows clarify allowable motion and diagnose missing constraints.
  • Status color-coding (under/over/fully constrained) provides at-a-glance quality checks.
  • Constraint graphs prevent hidden cycles that would otherwise only surface during regeneration.

Semantics that ride with the model: GD&T, PMI, and material intent

Geometric shape alone is mute about tolerances, datums, and function. Model-Based Definition (MBD) makes these semantics explicit by attaching Product Manufacturing Information (PMI) directly to faces, edges, and features. Datums display as annotated frames aligned to functional faces; feature control frames encode flatness, position, and profile; and materials, finishes, and process notes live as persistent attributes. This shift—formalized by ASME Y14.41 and ISO 16792—turned models into documents, where presentation views are curated: one view set shows primary datum structure, another focuses on hole callouts, a third on assembly-level tolerances. Importantly, these are not mere annotations; they are machine-readable semantics that downstream CMM software and CAM post-processors can consume to drive probing strategies, tool selection, and inspection plans.

Modern systems also manage model states—alternates that capture bend reliefs versus flat patterns, simplified reps, or envelope conditions—while binding them to PMI sets so that each state shows only its relevant definition. Materials and finishes are no longer just appearance textures; they carry density for mass properties, thermal coefficients for simulation presets, and notes that feed into ERP. When a face gets tagged as “datum A,” its shading can subtly change in PMI views, reinforcing its primacy in the tolerance stack. Tools such as PTC Creo, Siemens NX, and Dassault CATIA enrich this with semantic checks: a position tolerance lacking a proper datum reference frame flags red, and a profile tolerance over a non-contiguous face warns about scope. In that world, the viewer becomes a validator, and PMI glyphs are as central to comprehension as edges and faces.

  • GD&T frames anchor to topology, preserving meaning through edits and across formats like STEP AP242.
  • Materials are tied to calculations (mass, inertia, FEA presets) and procurement codes.
  • Saved MBD view sets act as ready-made communication packets for manufacturing and QA.

Behavior and relationships: kinematics, options, and embedded rules

Intent thrives in motion. Assembly kinematics visualize how mechanisms move across their envelopes, revealing interferences, over-travel, and required clearances. Exploded views, with trail lines drawn from part to nest, tell disassembly and assembly stories; explosion vectors show sequence and direction, while exploded line styles differentiate fasteners, covers, and consumables. Configurations and options—whether managed in CATIA Product Configurator, Siemens Knowledge Fusion, or PTC Creo Options Modeler—materialize as filterable trees and visual toggles that show which parts are common, optional, or mutually exclusive. Here, graphics and rules cooperate: selecting an option triggers a silent knowledge-based check that might color a part if a clash would result, or disallow a selection when design rules disapprove.

The landscape extends into explicit design rules. Knowledgeware and KBE systems attach checks such as “bolt protrusion length between 1–3 threads” or “minimum fillet vs. wall thickness” and then visualize results as on-model callouts or red/green overlays. Options models further rely on effectivity tags so reviewers can visualize different configurations by serial, market, or date. Breadcrumbs and mate graphs help navigate deep assemblies so reviewers always know where they are. Resultant visuals—motion envelopes, clearance halos, and effectivity filters—don’t just look helpful; they encode the product family’s logic, making intent about variation and behavior just as present on screen as shape.

  • Exploded views with trail lines communicate assembly sequences and service access.
  • Motion envelopes and collision halos visualize allowable travel and safety margins.
  • Option filters and effectivity layers make product-line logic visible and reviewable.

Analysis overlays that expose design rationale

Few visuals are as revealing as analysis maps. Curvature combs and zebra stripes illuminate the continuity and fairness of surfaces, crucial in Class-A surfacing with tools like Alias, CATIA, and NX. Draft analysis colorizes faces by angle relative to a pull direction, showing where injection molding might stick; thickness analysis shades thin sections that invite sink or structural weakness. In assemblies, interference and clearance are rendered as heatmaps, with legends indicating penetration depth or gap, while section cuts and clipping planes slice away occlusion to show internal fit and communication passages. These overlays grant reviewers an x-ray into the designer’s rationale, translating calculations into legible, navigable color and line.

CAE results mapped directly to CAD further tighten intent. Stress and strain contours from Abaqus, Simulia, or ANSYS overlay on bodies with synchronized units and legend scaling, while probe glyphs record sample points and min/max locations. CFD contours add pressure and velocity fields, often with streamlines that inherit color and thickness from scalar ranges. These are not mere pictures—metadata rides along: load cases, boundary conditions, and solver versions present as side panels, and hyperlinks back to the original setup ensure traceability. As review culture moved to DMU and beyond, the discipline shifted from reading plots in isolation to reading them in context of geometry and PMI, aligning numbers, tolerances, and manufacturability into one visual narrative.

  • Curvature, draft, and thickness maps transform geometric risk into immediate, spatial signals.
  • CAE contours alongside PMI let reviewers align performance with tolerances and processes.
  • Section cuts and clipping planes refine focus, letting teams interrogate interiors without remodels.

Communicating for downstream: view sets, lightweight, and DMU

Communication is where intent becomes shared reality. MBD view sets bundle zoom, orientation, visible items, section states, and PMI layers into named “slides,” each a curated message for manufacturing, inspection, or suppliers. Visualization platforms cache these states, allowing SMEs to flip between them without risk of misinterpretation. For large programs, lightweight visualizations are indispensable: formats like JT and 3D PDF (PRC-based) deliver tessellations with attached PMI and properties, enabling digital mock-up review without the full kernel present. In-browser DMU systems present explosion states, section managers, and measurement tools with enough fidelity to catch clashes, rule violations, and missing annotations, while remaining accessible to non-CAD stakeholders.

This downstream focus pushes clarity into the UI: PMI presentations can be sequenced, dimensions grouped by tolerance class, and measurement tools respect semantic edges rather than tessellated approximations. On the administrative side, release states and effectivity are filterable overlays, so reviewers see only what’s released for their purpose. The key is that the visualization becomes not simply an export but the primary actor—a delivery of portable intent rather than a screenshot. When the viewer encodes units, tolerances, and references unambiguously, the handoff across company lines can retain the reasons behind choices, not just the shapes themselves.

  • MBD view sets provide repeatable, authoritative snapshots of manufacturing intent.
  • JT and PRC-backed 3D PDFs distribute PMI-rich content to users without CAD seats.
  • DMU viewers preserve measurement accuracy and section semantics while scaling to massive assemblies.

Milestones: from pixels to semantics

1960s–1970s: from wireframe to shaded understanding

The earliest CRT displays showed intent in negative space: wireframes that flickered during rotations, challenging perception. Hidden-line removal research by Lawrence Roberts and Arthur Appel introduced algorithms that began to separate front from back, an early nod toward legibility rather than mere depiction. Ed Catmull’s invention of the Z-buffer provided a robust, real-time strategy to resolve visibility, which—paired with Henri Gouraud’s smooth shading and Bui Tuong Phong’s specular highlights—moved screens from mathematical sketches to visually continuous surfaces. Even then, however, most intent remained on paper drawings; the screen was a computational assistant, not the primary document.

Yet seeds were planted. Interactive editors started to show axes, grids, and tentative snap points, and cursor crosshairs became conversational: where you hovered mattered. Early systems at universities and labs experimented with digitizers and light pens, spawning the first glyphs for constraints like perpendicularity or tangency, if only as temporary guides. Although scarce, these cues foreshadowed a culture where graphics would teach users the model. In that phase, computational resource limits meant that any additional overlay carried cost, making each visual hint precious—and motivating the discipline of economical, meaning-laden graphics that would mature in decades to come.

  • Hidden-line and hidden-surface work (Roberts; Appel) elevated clarity over raw geometry.
  • Z-buffer plus Gouraud/Phong shading gave continuous cues for curvature and edges.
  • On-screen hints (axes, grids) seeded the idea that visuals could instruct as much as display.

1980s: solids arrive, features speak

The 1980s turned the abstract into tangible solids. Boundary representation (B-rep) left academia and entered commerce through Shape Data’s work that would underpin Parasolid, and Spatial Technology’s ACIS kernel founded by Dick Sowar. Ian Braid’s contributions to solid modeling theory helped formalize how faces, edges, and vertices cohered, granting kernels the rigor to support precise operations like Boolean, fillets, and shells. Visualization tracked these advances: edges differentiated between smooth and sharp, silhouette edges gained rendering prominence, and shading became accurate enough to support design judgements. But the truly transformative visual change came from feature-based, history-driven parametrics introduced commercially with PTC’s Pro/ENGINEER in 1987, envisioned by Sam Geisberg.

In Pro/E, sketches, dimensions, and features coalesced into a grammar where edits cascaded predictably—a dramatic leap from direct coordinate edits. Users watched as a modified dimension live-previewed an extrude update, with the regeneration navigator exposing failures. The feature tree and on-model dimensions taught users what to change and how the model would respond, a pedagogical loop embedded in the UI. Competitors and later entrants adopted and refined these patterns, and by the end of the decade the notion that a screen should expose intent—via dimensions, references, and rebuild order—had become part of the discipline’s DNA, setting the table for larger assemblies and rule-driven design.

  • Commercial B-rep kernels (Parasolid; ACIS) standardized precise, editable solids.
  • Feature-based parametrics made sketches and driving dimensions central to the visual experience.
  • Edge classification and shading improvements supported feature recognition by eye.

1990s: assemblies scale, constraints standardize, DMU grows

The 1990s were about scale and discipline. D-Cubed, under John Owen, created 2D/3D constraint solvers licensed by Unigraphics, SolidWorks, Autodesk, and others, standardizing glyphs, constraint behaviors, and diagnostics across the market. As mechanical design leapt to large assemblies, visual organization grew critical: sectioning tools matured, interference colorization highlighted clashes, and exploded views with trail lines taught assembly logic at a glance. Simultaneously, automotive surfacing tools like I-DEAS, CATIA, and Alias advanced curvature analysis—combs, porcupines, and zebra stripes—so surface quality could be assessed visually, aligning perception and math.

Digital Mock-Up (DMU) crystallized as a practice. Lightweight viewing became a necessity, with UGS/Unigraphics promoting the JT format to allow reviewers to load massive models, perform measurements, and analyze collisions without bringing full kernels to bear. UI conventions such as triad manipulators, isolate/show states, and transparency sliders let reviewers stage conversations: what-if motions, troubleshooting, and manufacturability discussions all flowed through a shared viewer. The decade ended with a clear principle: if a model’s behavior, tolerances, or fit matter, they must be seen—not just inferred—through well-designed visual encodings, accessible at scale.

  • D-Cubed solvers unified constraint visualization and editing patterns across platforms.
  • DMU practices formed with sectioning, clash color maps, and exploded review.
  • JT enabled lightweight, accurate measurements and review of large assemblies.

2000s: MBD, standards, and industrialized visualization

The 2000s codified semantics. ASME Y14.41 (2003) and ISO 16792 defined 3D PMI presentation, moving drafting standards into full 3D models. STEP AP203e2 and AP214 began to carry annotation payloads, converging toward what would become AP242’s semantic PMI. Vendors responded by streamlining PMI creation and display: instant dimensions, dynamic highlight, and clear datum callouts made it feasible to create fully toleranced models without 2D drawings. Visualization cores also matured: Tech Soft 3D’s HOOPS Visualize and SGI Open Inventor derivatives industrialized hidden-line rendering, sectioning, and on-model glyphs, letting OEMs and ISVs embed robust, consistent PMI overlays and HLR/HLV performance into their apps.

With MBD, the viewer became as important as the modeler. PMI view sets, filters, and validation tools appeared so teams could author and then verify that a model satisfied presentation and semantic rules. The decade also solidified the pipeline between CAD and CAE/CAM/Metrology: PMI fed CAM tolerances and CMM planning, with systems increasingly verifying that semantic tags survived across exchange formats. The concept of intent portability—ensuring color, glyphs, and semantics travel—shifted from aspiration to checklist, laying groundwork for the cloud and for the universal access culture that would dominate the following decade.

  • MBD gained formal standards and pervasive UI support for PMI creation and display.
  • Visualization toolkits (HOOPS, Open Inventor lineage) standardized glyphs and sectioning behavior.
  • STEP enhancements pushed toward semantic PMI interoperability.

2010s–2020s: real-time, cloud-native, and semantic delivery

A pair of revolutions defined the 2010s and early 2020s. First, cloud-native CAD and DMU arrived via WebGL—Onshape and Autodesk Fusion 360 placed assemblies, sections, and PMI-rich views in browsers, scaling collaboration without installs. Second, GPU ray tracing reached mainstream CAD: Siemens NX Ray Traced Studio, SolidWorks Visualize, and CATIA Live Rendering brought photoreal lighting into engineering contexts, while simultaneously offering non-photoreal “tech” modes (silhouettes, edge emphasis) optimized for manufacturing clarity. Exchange solidified around semantic PMI in STEP AP242; 3D PDF (PRC) and JT matured as standardized, lightweight carriers with rich PMI.

Parallel developments in BIM reinforced the trend: Autodesk Navisworks and Solibri popularized visual-first clash detection, review issues tracking, and rule checks, reminding the mechanical side that intent must be seen to be trusted. The cloud also reshaped scale: server-assisted rendering, tiled streaming, and GPU-backed web viewers made hundred-million facet reviews practical. The final click was cultural: teams began to expect that design rules, options, and tolerances would be as visible and filterable as geometry, and that a browser could show intent as faithfully as a desktop kernel—an expectation that today’s stacks largely meet.

  • WebGL-powered viewers enabled PMI, sectioning, and measure in the browser.
  • GPU ray tracing coexisted with NPR/technical modes to serve both persuasion and manufacturing.
  • AP242, JT, and PRC-backed 3D PDFs standardized lightweight, semantics-rich delivery.

How today’s stacks render intent

Graphics and kernels working in concert

Modern CAD viewers juggle two realities: a precise B-rep/NURBS world and a tessellated world optimized for interactive rendering. The kernel—Parasolid, ACIS, or an in-house engine—owns exact topology and curvature; the GPU consumes triangle meshes and per-vertex attributes. Intentful rendering demands a handshake between them. Kernel-driven hidden-line removal (HLR) produces watertight edge sets for technical illustration, while GPU HLR uses depth-peeling or stencil tricks for speed; hybrid systems select per view. Topology-aware selection and highlighting use kernel IDs to light faces, edges, and features consistently across LODs, ensuring a hole feature highlights as a unit even when triangle density changes.

Common toolkits and frameworks accelerate this integration. Tech Soft 3D’s HOOPS Visualize provides PMI glyphs, sectioning, and HLR primitives tuned for engineering semantics. Open Inventor’s lineage (via TGS and derivative SDKs) seeded scene graph patterns—state stacks, manipulators, and clipping nodes—still echoed in modern engines. Intel’s OSPRay delivers CPU ray tracing for accurate shadows and AO in massive models without dedicated GPUs, while the web rides on WebGL and looks toward WebGPU for compute-backed picking, tessellation refinement, and post-processing like edge enhancement. Across stacks, success hinges on consistent identifiers and metadata channels so that kernels can dictate smart highlighting, and viewers can carry PMI and constraint glyphs through culling and instancing without losing meaning.

  • Kernel IDs anchor topology-aware selection and avoid LOD-related highlight glitches.
  • Hybrid HLR strategies balance exactness (kernel) and interactivity (GPU).
  • OSPRay, HOOPS, and WebGL/WebGPU provide the rendering primitives for engineering semantics.

UI patterns that teach the model

Teaching through visuals requires consistent, low-friction patterns. Constraint and DOF color schemes communicate health; datum planes and axes render as translucent, branded overlays that can be toggled globally; manipulators and triads provide discoverable, constrained movement handles with snapping feedback. Mate graphs and breadcrumbs stay near the selection focus so users can climb or descend assembly hierarchies without losing context. Saved states—explodes, sections, isolate sets, and PMI presentations—live as first-class citizens, previewed with thumbnails that show section planes and active annotation groups.

Revision awareness is likewise visual. Model-compare modes colorize add/remove/modified faces; effectivity and options filters gray or hide out-of-scope parts; and rule checks paint violations in conspicuous shades, often paired with on-model badges that can be clicked to see the governing rule. Measurement tools respect semantic edges and show GD&T references in readouts; unit dropdowns and precision settings persist per view. The pattern is consistent: present state, afford action, and reinforce correctness. UI that does so turns the CAD scene into a living textbook where the model’s grammar—constraints, references, options, tolerances—is learned by sight and touch rather than by reading a manual.

  • Constraint/DOF colors and datum overlays make reference frames and health immediate.
  • Saved “view states” bind sections, explodes, and PMI into repeatable communication packets.
  • Model-compare and effectivity filters visualize change and configuration without ambiguity.

Analysis as intent revelation

Analysis visualizations act as truth serum for design choices. Curvature maps and zebra stripes double as training wheels, teaching designers what tangent and curvature continuity feel like; draft and thickness maps enforce manufacturability in molding and casting; and clearance bands show where assembly tolerance stacks demand attention. Embedded CAE overlays push further: stress/strain contours, mode shapes, frequency response plots, and CFD scalar/vector fields are not just colored skins but context-laden artifacts that include legends, units, min/max pointers, and probe glyphs. With tools from Abaqus/Simulia and ANSYS embedded in or adjacent to CAD, a reviewer can interrogate a hotspot, read permissible material stress from the property sheet, and verify that PMI tolerances are aligned with the predicted variability.

Clarity demands control. Legend scaling must be interactive; units must be explicit; cut planes should snap to datums; and probes must annotate values in editor-friendly formats for reports. Non-photorealistic rendering helps communicate—silhouette emphasis, hidden-line monochrome, and reduced shading draw attention to shape and annotation rather than to marketing gloss. While photorealism persuades, NPR conveys, and the best stacks let teams move between them without losing PMI, sections, or color encodings. Analysis, in this sense, becomes the connective tissue that binds intent to evidence, rendered in a way that anyone in the room can parse.

  • Curvature/draft/thickness overlays convert rules of thumb into visible thresholds.
  • CAE contours with probe glyphs and legends make results auditable and comparable.
  • NPR “tech” modes prioritize comprehension for manufacturing and inspection.

Data and standards that preserve meaning

Portable intent depends on standards. PMI/MBD rides on ASME Y14.41 and ISO 16792 for presentation, while STEP AP242 encodes semantic PMI such that datums, tolerances, and measurement references survive across vendors. The metrology world contributes QIF (Quality Information Framework), connecting nominal definitions to inspection plans and results. Lightweight and exchange formats—JT, 3DXML, PRC/U3D in 3D PDF, and glTF for the web—bind tessellation to metadata so that a measure tool hits the right face and PMI view sets reproduce as authored. Configurators and options integrate via KBE engines like Siemens Knowledge Fusion and PTC Creo Options Modeler, which inject rule evaluation into viewers so that invalid combinations can be flagged visually.

Constraint engines cement consistency. Siemens’ D-Cubed DCM/3D DCM underpin sketching and mating behavior across many platforms, ensuring stable glyphs, DOF logic, and solve diagnostics that users implicitly learn once and reuse everywhere. LEDAS heritage solvers power constraints in systems like BricsCAD, adding breadth to the ecosystem. These engines define not only mathematics but also visible semantics: which icon denotes parallelism, how over-constraints are displayed, and how failures are described. When data models, graphics stacks, and solvers agree on identifiers and semantics, visualization can truthfully mirror design meaning, making collaboration resilient across tools, teams, and time.

  • STEP AP242 and QIF align design, manufacturing, and inspection around semantic PMI.
  • JT, 3DXML, PRC/U3D, and glTF deliver lightweight yet meaning-rich review assets.
  • D-Cubed and LEDAS-derived solvers standardize constraint behavior and glyphs.

Conclusion

Where visualization and intent converge next

CAD’s journey has been from depicting shape to communicating design semantics: constraints, tolerances, behaviors, and manufacturability are now first-class visuals, not footnotes. Standards like ASME Y14.41, ISO 16792, and STEP AP242 made intent portable; constraint solvers and analysis overlays made it legible; and modern GPUs, ray tracing, and cloud viewers made it ubiquitous. The next frontier is deeper fusion—renderers that are natively aware of semantics, AI that summarizes “what matters” in a view, and AR/VR collaboration where PMI, rules, and analysis float in space with the model, validated against governing standards. Expect viewers to audit as they display: flagging nonconforming GD&T, revealing vanishing DOF as assemblies grow, and contrasting options configurations with auto-generated intent summaries that call out the deltas and their ramifications.

We will also see tighter contracts between kernels and graphics. WebGPU and compute-backed pipelines will enable on-the-fly re-tessellation to preserve PMI callout clarity at any zoom, topology-guided edge stylization to emphasize functionally critical boundaries, and per-pixel selection that honors precise edges without sacrificing interactivity. In tandem, exchanges will carry richer provenance—rule IDs, simulation case hashes, and inspection outcomes—so that any stakeholder viewing a model can trace what they’re seeing to the decisions and validations that justify it. In that world, “what you see is what you mean to build” becomes enforceable rather than aspirational, and visualization graduates from accompaniment to ground truth for engineering intent.

  • Semantic-aware rendering will preserve annotation clarity and precision at any scale.
  • AI-assisted views will highlight risk, change, and conformance to standards automatically.
  • AR/VR collaboration will merge spatial context with portable intent for instant shared understanding.



Also in Design News