---

Guy and Eitan discuss several very interesting new features recently announced by Microsoft.

Relevant links for more info:

• Stop defragmenting and start living: introducing auto index compaction | Microsoft Community Hub
• Automatic Index Compaction - SQL Server | Microsoft Learn
• Database DevOps (preview) in SSMS 22.4.1 | Microsoft Community Hub
• Manage SQL database schemas in VS Code: Publish dialog and item templates - Azure SQL Dev Corner
• Publish SQL Database Projects from Visual Studio Code - SQL Server | Microsoft Learn
• SQL code analysis in VS Code: Configure rules without editing your project file - Azure SQL Dev Corner
• SQL Projects Roadmap - DacFx Wiki
• Author Custom Code Analysis Rules - SQL Server | Microsoft Learn

Download audio: https://traffic.libsyn.com/secure/madeirasqlserverradio/SQLServerRadio_Show186.mp3?dest-id=213904

2026-04-13

• Install MCP servers from the registry with guided configuration directly in the CLI
• Esc key works correctly after a failed /resume session lookup
• Persist resolved model in session history and defer model changes during active turns
• ACP clients can now provide MCP servers (stdio, HTTP, SSE) when starting or loading sessions
• The --config-dir flag is now respected when selecting the active model
• Add /env command to show loaded environment details (instructions, MCP servers, skills, agents, plugins)
• /share appends the correct file extension (.md or .html) when a custom output path is given without one
• /add-dir accepts relative paths (e.g. ./src, ../sibling) and resolves them to absolute paths
• Custom instruction files preserve special characters like & and
• Skill picker list scrolls correctly when the list exceeds the terminal height
• MCP client reports the correct CLI version during server handshake
• /logout shows a warning when signed in via gh CLI, PAT, API key, or environment variable, since /logout only manages OAuth sessions
• Alt+D now deletes the word in front of the cursor in text input
• /share html shows a file:// URL and supports Ctrl+X O to open the file directly
• Skill instructions persist correctly across conversation turns
• You can now remote control your CLI sessions using --remote or /remote
• MCP remote server connections automatically retry on transient network failures
• Share Research TOC sidebar anchor links navigate correctly within the page

Roslyn analyzers are great for enforcing coding standards and finding bugs. However, they can sometimes report false positives or warnings you want to ignore in specific contexts. You can suppress these warnings using #pragma directives or [SuppressMessage] attributes, but both approaches require modifying the source code. A DiagnosticSuppressor is a Roslyn component that suppresses diagnostics reported…

Discover the latest ServerTextControl skills for AI-assisted document processing, enhancing your Text Control Agent capabilities. Learn how these new skills can streamline your workflow and improve efficiency in handling documents.

## 1 Normalization vs. Reality: When to Denormalize for Throughput on SQL Server & Azure SQL

### 1.1 The Architect’s Paradox: Purity vs. Performance

Normalization still matters. Third Normal Form and BCNF are still the right tools when the goal is to keep data clean, reduce duplication, and avoid update anomalies. In systems of record, those are not textbook concerns. They affect billing, inventory, approvals, reporting, and every cleanup script someone has to write after bad data gets into production.

But there is another side to this. The schema that is best for correctness is not always the schema that is best for throughput. That gap gets wider when the system becomes read heavy, runs at scale, and serves the same data shape over and over. A fully normalized design can be logically correct and still expensive to query under load.

In SQL Server and Azure SQL, denormalization should not be the first response to slow queries. Many systems have more basic issues first: weak indexing, bad query patterns, parameter-sensitive plans, ORM chatter, and avoidable key lookups. Those problems need to be fixed before the schema becomes the main suspect. But once those are under control, some workloads still hit a wall because the engine spends too much time assembling data across too many tables.