Add VERSION2.md summarizing key features for 2.0 pre-release

Documents major features:
- Redesigned three-layer type system
- Object storage (Augmented Schema)
- Autopopulate 2.0 jobs system
- Semantic matching for joins
- Simplified settings management
- Python 3.10+ and MySQL 8.0+ requirements

Co-authored-by: dimitri-yatsenko <dimitri@datajoint.com>

Author: claude

VERSION2.md

@@ -0,0 +1,152 @@
+# DataJoint 2.0 Pre-Release Notes
+
+This document summarizes the key features and changes in DataJoint 2.0.
+
+## Major Features
+
+### 1. Redesigned Type System
+
+A three-layer type architecture provides clarity and flexibility:
+
+- **Layer 1 - Native Database Types**: Backend-specific types (`FLOAT`, `TINYINT UNSIGNED`, `LONGBLOB`)
+- **Layer 2 - Core DataJoint Types**: Scientist-friendly, standardized across backends (`float32`, `uint8`, `bool`, `json`, `int64`, `uint64`)
+- **Layer 3 - AttributeTypes**: Composable types with `encode()`/`decode()` semantics using angle bracket syntax (`<djblob>`, `<object>`, `<content>`)
+
+**Key changes:**
+- `blob` now stores raw bytes; use `<djblob>` for serialized Python objects
+- New type aliases: `int8`, `int16`, `int32`, `int64`, `uint8`, `uint16`, `uint32`, `uint64`, `float32`, `float64`, `bool`
+- AttributeTypes are composable: `<xblob>` wraps `<content>` which wraps `<djblob>`
+
+### 2. Object Storage (Augmented Schema)
+
+The new `object` type provides managed file/folder storage tightly coupled with the database:
+
+```python
+@schema
+class Recording(dj.Manual):
+    definition = """
+    subject_id : int
+    session_id : int
+    ---
+    raw_data : object              # uses default store
+    published : object@public      # uses named store
+    """
+```
+
+**Features:**
+- Supports both files and folders (including Zarr, TileDB)
+- Two insert modes: copy (small files) and staged (large objects written directly to storage)
+- Automatic deletion when records are deleted
+- Content-addressed storage for deduplication (`<content>` type)
+- Path-addressed storage for primary-key based organization (`<object>` type)
+
+### 3. Autopopulate 2.0 Jobs System
+
+Per-table job management replaces the schema-wide `~jobs` table:
+
+```python
+# Access jobs for a specific table
+FilteredImage.jobs
+FilteredImage.jobs & 'status="error"'
+FilteredImage.jobs.refresh()
+```
+
+**Improvements:**
+- Rich status tracking: `pending`, `reserved`, `success`, `error`, `ignore`
+- Per-table jobs tables with full primary key visibility (no key hashing)
+- Priority scheduling and delayed execution support
+- Worker tracking: host, PID, connection ID, version
+- Automatic stale job cleanup via `refresh()`
+
+### 4. Semantic Matching for Joins
+
+Joins now use **lineage-based matching** instead of pure name matching:
+
+- Attributes are only matched when they share both the same name AND the same origin
+- Prevents accidental joins on unrelated attributes that happen to share names
+- Enables valid joins that were previously blocked
+- Lineage tracked via hidden `~lineage` table per schema
+
+**Example:**
+```python
+# Old behavior: Would incorrectly join on 'id'
+Student * Course  # ERROR: Student.id and Course.id have different lineage
+
+# New behavior: Joins proceed when lineage matches
+FavoriteCourse * DependentCourse  # OK: both course_id trace to Course.course_id
+```
+
+### 5. Simplified Settings Management
+
+- Pure Pydantic-based configuration
+- Recursive config file search (`datajoint.json`)
+- Secrets separation support
+- Removed deprecated `save()` methods
+
+### 6. Modern Python and MySQL Support
+
+- **Python 3.10+** required (dropped support for earlier versions)
+- **MySQL 8.0+** required (dropped support for earlier versions)
+- Source layout (`src/datajoint/`)
+- Pre-commit hooks with ruff linting and formatting
+
+## Breaking Changes
+
+### Removed Features
+
+- **`~log` table**: Use Python logging instead
+- **Legacy `AttributeAdapter`**: Replaced by `AttributeType` system
+- **`blob` serialization**: `blob`/`longblob` now store raw bytes; use `<djblob>` for serialized objects
+- **Schema-level jobs table**: Replaced by per-table jobs (`MyTable.jobs`)
+- **`AutoPopulate.target` property**: Removed
+- **`AutoPopulate._make_tuples`**: Use `make()` method exclusively
+- **`dj.config.save()` methods**: Configuration is now read-only at runtime
+
+### API Changes
+
+- `dj.U` join is deprecated (use restriction instead)
+- Jobs API moved from `schema.jobs` to `Table.jobs`
+- External storage requires explicit store configuration
+
+## Migration Notes
+
+### Blob Columns
+
+Existing `longblob` columns with serialized data should be migrated to `<djblob>`:
+
+```python
+from datajoint.migrate import migrate_blob_columns
+migrate_blob_columns(schema)
+```
+
+### Jobs Tables
+
+The schema-wide `~jobs` table is no longer used. Jobs are now per-table and automatically created when needed. Old `~jobs` data can be manually migrated if historical error logs are needed.
+
+## Configuration
+
+Example `datajoint.json`:
+
+```json
+{
+  "database.host": "localhost",
+  "database.user": "datajoint",
+  "database.password": "password",
+  "stores": {
+    "main": {
+      "protocol": "s3",
+      "endpoint": "s3.amazonaws.com",
+      "bucket": "my-bucket",
+      "location": "datajoint-store"
+    }
+  }
+}
+```
+
+## Version
+
+```
+2.0.0a0
+```
+
+This is an alpha pre-release for testing and integration purposes.