CHANGING THE CLIENT, KEEPING THE WORLD
By Jeromy Stroh, Pixel Games LLC R
eplacing a live game client is a different
problem from building a new game. The interface can change completely, but the world still belongs to people who have invested time in it. Their characters, equipment and familiar actions matter more than the novelty of the replacement. Fantasy Online 2’s PxEngine launch provides a useful case. The
replacement reached the live game in August 2026 after more than six months of development. Its lessons extend beyond choosing a renderer: compatibility, presentation and deployment all become part of the player’s experience of continuity. One early constraint was text. The previous interface used BabylonJS
GUI and a Canvas 2D rendering path. Changing text meant repeated CPU rasterization. Bitmap fonts reduced that work, but required separate atlases for different sizes and introduced compromises when scaled. The early PxEngine work moved toward GPU text using multi-
channel signed distance fields. Glyph geometry and layout could be prepared when the text changed, then batched for rendering. The broader lesson is to identify the repeated work behind a symptom. Replacing a visible widget without changing that work can preserve the original bottleneck. That principle also applies to networking. A client needs enough
information to present the world, but it does not need every static definition repeated whenever a small piece of state changes. Published project notes describe separating definition catalogs from ordinary gameplay updates, while the server retained authority over combat, quest progress and rewards. The important distinction is between a catalog and an individual
player’s state. An item’s description is different information from who owns it. Separating those responsibilities does not make the interface authoritative. It makes the contract clearer: presentation can change while the rules governing ownership, progress and valid actions remain consistent. Compatibility therefore deserves its own acceptance criteria. The
launch notes describe smaller player-stat updates that preserved support for the legacy client. That is more useful than treating a
22 | MCV/DEVELOP November/December 2026
successful connection as proof that two clients agree. A message can arrive correctly while a party meter, selected target or cooldown display becomes misleading. Presentation introduces another boundary. A renderer can produce
frames regularly while the native window presents them unevenly. The published Steam investigation distinguished internal rendering from the later texture handoff. At publication, the repair was implemented and had passed checks, but awaited the next Steam build. This illustrates why implementation and release are separate claims. Input demands similar care. Published patch notes record fixes
including clicks near the HUD moving the character and focused controls intercepting gameplay shortcuts. These are small-looking defects with large consequences: players experience them as unreliable actions. Testing isolated controls misses the transitions between a text field, a window and the surrounding game. A useful migration checklist follows those transitions. What happens
when a window opens during movement? When focus leaves and returns? When a viewport changes size? When an update arrives while an
interface is closed? These questions are less glamorous than a new rendering effect, but closer to the failures that interrupt play. Finally, launch is a change in responsibility, not the end of the project.
Once the replacement becomes the normal client, feedback comes from a wider range of screens, keyboards and habits. The public rollout continued with bug reporting, compatibility work and platform-specific fixes. The practical measure of a replacement is not how different it looks. It is whether the new client can keep evolving while the player’s
relationship with the existing world remains understandable and dependable. Keeping those
boundaries explicit also makes release claims safer: a prototype, a tested candidate and a production deployment describe different stages of the development and release process.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52