Release target patterns

Dev-Centr names sets of release triplets (OS × CPU) the way Gentoo names Portage profiles: a short word for intent, plus an optional year when the set must not drift. Agents expand these names from patterns.sdl (skill release-targets). CI wiring: Polyglot CI matrix.

flowchart TB
  subgraph living["Living names → current_year"]
    common["common / default"]
    desktop["desktop"]
    cloud["cloud"]
    lab["lab / hpc"]
    edge["edge"]
    omnibus["omnibus / all"]
  end
  common --> winlin["win+lin × x64+arm64"]
  desktop --> common
  desktop --> mac["macos-arm64"]
  cloud --> lin["linux × x64+arm64"]
  lab --> lin
  lab --> hpcisa["ppc64le + riscv64 + s390x"]
  edge --> armrv["linux-arm64 + riscv64"]
  omnibus --> everything["entire catalog"]
  dated["common/2026"] -.->|"freeze"| common

Why name the sets

Without names, every repo invents a matrix. With names, “ship common” means the same four Windows/Linux slices house-wide, and “ship common/2025” reconstructs what “default” meant in an older publication after the living default moves.

Nomenclature

Pattern tokens

Pattern Alias Intent (2026)

common

default

Mass-market floor: Windows + Linux, x64 and arm64 (four triplets).

desktop

—

Gentoo-style desktop overlay: common plus macOS arm64 (Apple Silicon). Prefer for GUI apps.

cloud

—

Linux x64 + Linux arm64 only (containers / VMs).

lab

hpc

Research / HPC: Linux x64 + arm64 + ppc64le + riscv64 + s390x.

edge

—

SBC / edge: Linux arm64 + Linux riscv64.

heritage

—

Museum / long-tail (32-bit x86, armv7). Document; do not build unless asked.

omnibus

all

Every triplet in the living catalog. Unusual — confirm before enabling.

Etymology is deliberate: Gentoo’s default/…/desktop hierarchy, the Victorian omnibus (“carries everyone”), and plain lab / edge for audience.

Dating and drift

Form Meaning

Bare name (common)

Latest definition under current_year in patterns.sdl.

common/2026 or “common 2026”

Frozen 2026 set — safe to cite from old READMEs and archaeology.

Reconstructing “default” on a 2025-era project

Treat as common/2025 (or ask), never silently upgrade to living common.

When a new ISA becomes house-default (for example POWER in a later year), bump current_year, keep the previous year block immutable, and teach agents to prefer dated citations in archaeology.

What to ship in 2026

Most consumers → common (GUI → desktop)

Suggestion for 2026 compatibility with most desktop/server consumers:

  • Required floor (common/2026): windows-x64, windows-arm64, linux-x64, linux-arm64.

  • GUI apps (desktop/2026): add macos-arm64. Do not schedule GitHub-hosted Intel Mac runners.

That matches dual-ISA Windows (Snapdragon / Windows on Arm), dual-ISA Linux (cloud Ampere/Graviton + classic x86_64), and Apple’s arm64-only Mac fleet.

High-performance researchers → lab

Add Linux ppc64le (IBM POWER), riscv64, and s390x (IBM Z) beside the Linux dual-ISA pair. Keep Windows/macOS slices only if the researchers also run those hosts for tooling — they are not part of lab itself.

Use omnibus only when you truly need heritage + BSD + everything in one release train.

Architecture usage statistics (cited)

Figures below are order-of-magnitude orientation, not a substitute for your product’s telemetry. Always prefer the primary survey page when making a go/no-go call; secondary articles paraphrase market firms that often paywall full tables.

Client CPU: x86 vs Arm (shipments)

Mercury Research’s Q2 2026 client estimates (reported by trade press) put Arm-based client CPUs around 15% of the client CPU market versus x86, with Apple and Windows-on-Arm among the drivers. AMD’s x86 share within the x86 segment reached record highs the same quarter (desktop x86 share reported near 35% for AMD vs Intel).

Sources
Implication for common/2026

Shipping both x64 and arm64 on Windows and Linux covers the dominant x86 install base and the growing Arm client minority without waiting for Arm to “win” gaming.

Gaming / Steam panel (architecture mix)

Valve’s opt-in Steam Hardware & Software Survey remains overwhelmingly 64-bit x86 for gaming PCs. Secondary reporting of the December 2025 survey wave cited roughly 97% x64 vs about 0.45% Arm among surveyed systems — useful as a lower bound for game reach, not for general productivity PCs or Macs.

Sources
Steam OS share (July 2026 reporting)

Windows ≈ 93.7%, Linux ≈ 4.0%, macOS ≈ 2.3% on the Steam panel (Command Linux summary of Valve’s July 2026 survey).

General desktop OS sessions

StatCounter’s public desktop OS tracker (July 2026 reporting) shows Windows still majority, with macOS and Linux as meaningful minorities — a reminder that Mac arm64 matters for GUI desktop even when Steam under-represents Mac.

Sources

Servers, cloud, and HPC

  • Cloud Linux arm64 (Graviton, Ampere, Cobalt, etc.) is first-class for common / cloud — treat Linux arm64 as required, not experimental.

  • TOP500 supercomputers are effectively all Linux; CPU mixes include x86_64 and other server ISAs over time — motivating lab’s ppc64le / s390x / riscv64 extras for research audiences rather than for Steam-like consumers.

  • CNCF cloud-native surveys consistently show Linux dominating production Kubernetes clusters (see CNCF annual survey publications; SQ Magazine summarizes recent waves).

How to read the table for shipping

Audience Pattern Stats takeaway

General desktop + server consumers

common / desktop

x86_64 still majority; Arm client ~mid-teens% and rising; Mac = arm64; Steam Arm still niche for games.

Cloud-native services

cloud

Dual-ISA Linux covers the hosts you actually deploy.

HPC / academic labs

lab

Add POWER, RISC-V, Z — low Steam share, high cluster relevance.

Exhaustive / compliance theater

omnibus

Stats do not justify this for most products — use sparingly.

Related