# WPMU Dev Replacement Project

**Status:** Phase 1 Complete (Analysis)  
**Cost Impact:** $298/year subscription savings  
**Effort Estimate:** 6-8 weeks for production-ready implementation  
**Team:** WordPress specialist + Elementor integration specialist  

---

## Project Goals

Replace WPMU Dev's Hummingbird (caching) and Smush (image optimization) plugins with custom-built solutions that are:

1. **Infrastructure-specific** — Optimized for your brisbane01 single-server setup, not generic WordPress
2. **Elementor-native** — Deep integration with Elementor's CSS caching and data storage
3. **Transparent** — Open-source, auditable code you fully control
4. **Cost-saving** — Eliminate $298/year WPMU Dev subscription immediately

---

## Repository Structure

```
wpmu-dev-analysis/
├── 01-AUDIT-FINDINGS.md              # Complete audit of current plugins
├── 02-TECHNICAL-SPECIFICATION.md     # Architecture, APIs, integration points
├── 03-PHASE-1-RESEARCH.md            # (Coming: benchmark data, detailed design)
├── 04-BUILD-LOGS.md                  # (Coming: development progress)
├── 05-DEPLOYMENT-RUNBOOK.md          # (Coming: step-by-step rollout)
├── plugins/                          # (Coming: actual plugin code)
│   ├── help4bis-performance-cache/
│   └── help4bis-image-optimizer/
└── tests/                            # (Coming: test suites)
```

---

## Key Findings

### Hummingbird Status: Mostly Disabled
- ✅ Browser cache headers set (1 year for CSS/JS/images)
- ❌ Page HTML caching: DISABLED
- ❌ Minification: DISABLED
- ❌ Lazy load: DISABLED
- ✅ Database cleanup: Runs weekly
- **Impact:** Minimal performance contribution. Can replace with ~400-line plugin.

### Smush Pro Status: Actively Working
- ✅ Auto-compress on upload
- ✅ WebP generation
- ✅ Lazy load integration
- ❌ Responsive image sizing: NOT DONE
- ❌ CDN delivery: NOT USED
- **Impact:** Real value. Image compression + WebP saves 40-60% bandwidth. Custom replacement is viable but requires ImageMagick/cwebp.

### Infrastructure Summary
| Metric | Value |
|--------|-------|
| Total WordPress sites | 14 |
| Using Elementor | 10 |
| Using Hummingbird | 3 |
| Using Smush Pro | 8 |
| Total image assets | 9GB |
| Active cache on disk | ~58MB |
| Subscription cost/year | $298 |

---

## Phase Timeline

### Phase 1: Research & Design ✅ DONE
- [x] Comprehensive audit of plugins
- [x] Configuration analysis
- [x] Technical specification
- [ ] Interview Elementor specialist (next)
- [ ] Create performance benchmarks (next)

### Phase 2: Build Cache Plugin (2 weeks, estimate 400-600 LOC)
- [ ] HTTP header logic
- [ ] Database cleanup
- [ ] Elementor CSS cache integration
- [ ] Admin settings UI
- [ ] Testing

### Phase 3: Build Image Optimizer (3 weeks, estimate 1000-1500 LOC)
- [ ] Upload handler + ImageMagick compression
- [ ] WebP generation
- [ ] Lazy load injection
- [ ] Responsive srcset generation
- [ ] Elementor integration
- [ ] Bulk optimizer for existing images
- [ ] Testing

### Phase 4: Production Rollout (1 week)
- [ ] Disable Hummingbird, test cache plugin (72 hours)
- [ ] Disable Smush, test image optimizer (72 hours)
- [ ] Roll out to all 14 sites
- [ ] Cancel WPMU Dev subscription

### Phase 5: Ongoing Maintenance
- [ ] Monitor image optimization logs
- [ ] Browser compatibility testing
- [ ] Quarterly security updates

---

## Critical Decisions Made

### 1. ✅ Skip Page HTML Caching
Hummingbird's page HTML caching is complex and brings little value for Elementor sites (which are inherently dynamic). We're implementing browser cache headers only.

**Trade-off:** ~3% potential speed gain (page HTML caching) vs. 40% reduced complexity. We choose simplicity.

### 2. ✅ Use Local ImageMagick Instead of Cloud
Smush uses WPMU Dev's cloud API for heavy lifting. We'll use ImageMagick (already on brisbane01).

**Trade-off:** Smush cloud handles edge cases better; local ImageMagick is 95% as good. We gain control, lose cloud failover.

### 3. ✅ Elementor-First Integration
Both plugins must have explicit Elementor hooks (CSS cache, image URL rewriting, lazy load in rendered output).

**Trade-off:** Elementor-specific code means less reusability to other page builders. But 10/14 sites use Elementor, so ROI is high.

### 4. ✅ Single-Server Assumption
No Redis, no Memcached, no distributed caching. brisbane01 is a single production server.

**Trade-off:** Can't scale to multiple servers without refactoring. But for current infrastructure, this is appropriate.

---

## Risk Mitigation

### Risk: Broken Images During ImageMagick Conversion
**Severity:** HIGH  
**Mitigation:** Graceful fallback — if ImageMagick fails, upload original image uncompressed. Log the error for admin review.

### Risk: Elementor CSS Cache Invalidation Bugs
**Severity:** HIGH  
**Mitigation:** Explicit hooks into Elementor's save actions. Comprehensive testing. Clear browser cache between deployments.

### Risk: WebP Browser Support
**Severity:** LOW  
**Mitigation:** Native `<picture>` tag with JPEG fallback. Supported in 95%+ of browsers. Old browsers just get JPEG.

### Risk: Performance Regression
**Severity:** MEDIUM  
**Mitigation:** Benchmark current sites before/after. Monitor Core Web Vitals for 2 weeks post-rollout.

---

## How to Use This Repository

### For Developers
1. Read `01-AUDIT-FINDINGS.md` for context
2. Read `02-TECHNICAL-SPECIFICATION.md` for implementation details
3. Start Phase 2 development in `plugins/help4bis-performance-cache/`
4. Document decisions in `04-BUILD-LOGS.md`

### For Project Managers
1. Check Phase Timeline above for progress
2. Monitor for blockers (Elementor specialist availability, ImageMagick verification)
3. Review performance benchmarks in Phase 1 Research (coming)

### For QA/Testing
1. Use test checklist in Technical Specification § Testing Strategy
2. Run on staging sites (v-i-o.com, stald.com.au for image heavy)
3. Document issues with detailed reproduction steps

---

## Next Steps

1. ✅ **Approve Phase 1 findings** — confirm scope is correct
2. ⏳ **Assign Elementor specialist** — 4 hours to review cache invalidation strategy
3. ⏳ **Benchmark baseline** — Run Google PageSpeed on 3-4 key sites (record before/after)
4. ⏳ **Verify ImageMagick** — Confirm `convert`, `magick`, `cwebp` are installed on brisbane01
5. ⏳ **Set up staging** — Mirror v-i-o.com (image-heavy) for Plugin testing
6. ⏳ **Start Phase 2** — Begin cache plugin development

---

## Assumptions & Dependencies

### Assumptions
- brisbane01 remains single-server (no scaling to multiple servers)
- ImageMagick/cwebp remain available and maintained
- Elementor API doesn't change significantly (Elementor < 4.0)
- Hummingbird deactivation doesn't break any custom code

### Dependencies
- ImageMagick (for image compression)
- cwebp (for WebP generation, optional but recommended)
- PHP >= 7.4 (WordPress minimum)
- WordPress >= 5.0
- Elementor (for 10/14 sites)

### External Teams
- **Elementor specialist** — consult on CSS cache invalidation (4 hours)
- **QA team** — test across browsers (Chrome, Firefox, Safari, Edge)

---

## Budget & ROI

| Category | Amount | Notes |
|----------|--------|-------|
| **Development** | 80-120 hours | At $50-100/hr = $4K-12K |
| **Testing** | 20-30 hours | Staging + production validation |
| **Documentation** | 10 hours | Build logs, runbooks |
| **Annual savings** | $298/year | WPMU Dev subscription |
| **Breakeven (financial)** | 13-53 years | Not the real ROI |
| **Real ROI** | Immediately | Control + transparency + knowledge |

---

## Questions & Clarifications

**Q: Why not just keep WPMU Dev?**  
A: You're paying $298/year for features you mostly don't use (Hummingbird is disabled, Smush is generic). Custom plugins give you control and save money.

**Q: Will custom plugins be as good as WPMU Dev?**  
A: Better for your infrastructure. Worse for edge cases. We trade WPMU Dev's generalist approach for your specific needs (Elementor integration, ImageMagick optimization).

**Q: What if we scale to multiple servers later?**  
A: Refactor to add Redis/memcached at that time. Current design doesn't preclude it; we just don't implement it now.

**Q: How do we rollback if something breaks?**  
A: Deactivate the custom plugin, reactivate Hummingbird/Smush. WPMU Dev is still licensed until we cancel. We keep them active during Phase 4 rollout.

---

## Contact & Escalation

- **Technical questions:** Claude Code (this project)
- **Elementor integration:** (assign specialist)
- **Performance concerns:** Claude Code (monitoring tools)
- **Subscription cancellation:** (after Phase 4 complete)

---

**Last updated:** 2026-05-17  
**Status:** Phase 1 complete, awaiting approval for Phase 2  
**Repository:** https://192.168.0.175:3100/help4bis/wpmu-dev-analysis