1. Forum
  2. DOVE-Net
  3. Synchronet Programming

  • src/doors/syncconquer/DESIGN.md

    From Rob Swindell (on Debian Linux)@VERT to Git commit to main/sbbs/master on Mon Jul 13 00:36:58 2026
    https://gitlab.synchro.net/main/sbbs/-/commit/a1610a2bae948446be72c7f6
    Modified Files:
    src/doors/syncconquer/DESIGN.md
    Log Message:
    syncconquer: full DESIGN.md Phase-2 dirty-rect spec

    Expand the brief Phase-2 note into a proper spec: one transport-agnostic framebuffer differ feeding a per-tier rect encoder (DrawJXL rects for JXL clients; a single reused sixel-rect backend for every non-JXL client, since
    APC PPM at CTerm 1316 is newer than Sixel at 1189 so PPM implies Sixel); coalescing + a whole-frame fallback threshold; a cached/sixel background
    (never raw PPM) with the C;S/C;L image-cache manager for static UI art; measured full-frame throughput anchors and a per-scene win table; the
    ~4% PPM-on-JXL break-even; SF #258 and map-scroll constraints; and a measurement-probe-first implementation order.

    Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

    ---
    þ Synchronet þ Vertrauen þ Home of Synchronet þ [vert/cvs/bbs].synchro.net
  • From Rob Swindell (on Debian Linux)@VERT to Git commit to main/sbbs/master on Tue Jul 14 21:04:15 2026
    https://gitlab.synchro.net/main/sbbs/-/commit/24ed0fe2c0a6cfe67c608656
    Modified Files:
    src/doors/syncconquer/DESIGN.md
    Log Message:
    syncconquer: correct DESIGN.md Phase-2 tier facts + measured throughput

    Reconcile the Phase-2 dirty-rect spec with how the door actually behaves and with real probe measurements:

    - The tier default is NOT a divergence from the FPS doors: sa_auto_tier()
    (JXL -> sixel -> PPM/text) is shared by all the doors. The real divergence
    is sixel handling -- RA/TD render native 640x400 and send sixel 1:1 with no
    reliance on client pan/pad=2 doubling, vs Duke/DOOM's half-res + client
    doubling. Server-side scale only for sub-640x400 or non-mod-6 height; width
    clamped to the graphics ceiling. Huge-canvas client doubling noted as a
    possible optimization, not "never".
    - Expected throughput replaced with measured per-tier numbers: JXL ~4x (the
    default tier, but area-scale-optimistic), sixel ~7x (trustworthy, fallback),
    text ~0.8x (regresses -- exclude). Near-static scenes ~167x.
    - Implementation order updated: the next probe refinement is to encode JXL
    rects for real (JXL being the default tier); sixel-rect is the biggest
    trustworthy win but serves the fallback; skip the already-diffed text tiers.

    Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

    ---
    þ Synchronet þ Vertrauen þ Home of Synchronet þ [vert/cvs/bbs].synchro.net