Architecture hinzugefügt

Architecture.md

@@ -0,0 +1,168 @@
+Willkommen im Wiki.# Architecture
+
+## High-Level Runtime Architecture
+
+Default Microsoft SQL Server mode:
+
+```text
++----------------------+
+|   Romexis Clients    |
++----------+-----------+
+           |
+           | RMI / Romexis protocol ports
+           v
++----------------------+
+| Romexis Server       |
+| Docker Container     |
++----------+-----------+
+           |
+           | JDBC
+           v
++----------------------+
+| Microsoft SQL Server |
+| Docker Container     |
++----------------------+
+```
+
+Firebird mode:
+
+```text
++----------------------+
+|   Romexis Clients    |
++----------+-----------+
+           |
+           | RMI / Romexis protocol ports
+           v
++----------------------+
+| Romexis Server       |
+| Docker Container     |
++----------+-----------+
+           |
+           | JDBC / Jaybird
+           v
++----------------------+
+| Firebird Server      |
+| Docker Container     |
++----------------------+
+```
+
+---
+
+## Image Architecture
+
+```text
+romexis-payload:<version>
+  /opt/romexis
+  /opt/romexis-mssql-db
+  /opt/romexis/version
+
+romexis-firebird-payload:latest
+  /opt/romexis-firebird-db
+  Firebird SQL scripts
+  Firebird backup/restore helpers
+  romexis_new.fdb template/reference
+
+romexis-base-jre:11-<arch>
+  Java 11
+  JavaFX/OpenJFX
+  database client tools
+  system dependencies
+
+romexis-server:<version>-<arch>
+  Romexis payload
+  Firebird payload
+  base runtime
+  Java property agent
+  initialization scripts
+  entrypoint
+```
+
+---
+
+## Why Payload Images?
+
+The original Dockerfile downloaded and extracted installer files inside the final server build.
+
+That worked, but had drawbacks:
+
+- repeated large downloads
+- slow builds
+- harder debugging
+- installer extraction tightly coupled to server image build
+- poor reuse across architectures
+
+Payload images solve this by making the extracted installer content a reusable build artifact.
+
+---
+
+## Runtime Data Layout
+
+The runtime stack uses persistent volumes or host directories for:
+
+```text
+/data/romexis_images
+/data/romexis_ergodata
+/data/romexis_cache
+/opt/romexis/sconfig
+/opt/romexis/programdata
+```
+
+Database data is backend-specific:
+
+```text
+Microsoft SQL Server:
+  /var/opt/mssql
+
+Firebird:
+  /firebird/data/romexis.fdb
+```
+
+---
+
+## Startup Flow
+
+```text
+entrypoint.sh
+  |
+  |-- prepare persistent directories
+  |-- generate database URL if needed
+  |-- write Romexis configuration
+  |-- run /opt/init-romexis-db.sh
+  |     |
+  |     |-- route to MSSQL init if SERVER_DB=5
+  |     |-- route to Firebird init if SERVER_DB=4
+  |
+  |-- fix keystore alias
+  |-- start Romexis Server
+```
+
+---
+
+## Database Initialization Router
+
+The runtime uses a small router script:
+
+```text
+/opt/init-romexis-db.sh
+```
+
+It detects the backend using:
+
+```text
+ROMEXIS_DB_URL
+SERVER_DB
+```
+
+Supported backend identifiers:
+
+| Value | Backend |
+|---|---|
+| `SERVER_DB=5` | Microsoft SQL Server |
+| `SERVER_DB=4` | Firebird |
+
+Backend-specific logic is split into:
+
+```text
+/opt/init-romexis-mssql-db.sh
+/opt/init-romexis-firebird-db.sh
+```