Skip to content

Latest commit

 

History

History
263 lines (177 loc) · 8.3 KB

File metadata and controls

263 lines (177 loc) · 8.3 KB

Research Papers for Janus-Datalog

Date: October 2025 Author: Wes Brown (with Claude Code assistance)


What This Is

This directory contains five research paper proposals and outlines documenting novel contributions from the Janus-Datalog project. These range from initial proposals (Papers 1-3) to detailed outlines ready for development (Papers 4-5).


The Files

📋 SUMMARY.md (14KB)

Start here. Executive summary of Papers 1-3 covering:

  • Why publish these papers
  • Comparison matrix
  • Prioritization recommendations
  • Collaboration options
  • Venue selection guidance
  • Template emails for reaching out

Paper Proposals (Initial Ideas)

Title: "When Greedy Beats Optimal: Join Ordering for Pattern-Based Datalog Queries Without Statistics"

Status: Initial proposal - needs correction (makes unsupported claims about beating cost-based)

See instead: STATISTICS_UNNECESSARY_PAPER_OUTLINE.md for corrected version

Title: "From Theory to Practice: Implementing Datalog as Pure Relational Algebra"

Key Result: Production Datalog using ONLY classical relational operations (π, σ, ⋈)

  • No semi-naive evaluation
  • No magic sets transformation
  • 13-87% faster than specialized strategies
  • 10× simpler implementation

Impact: HIGH - closes 40-year theory-practice gap Target: SIGMOD/VLDB/PODS

Title: "From Volcano to Lazy Sequences: Functional Streaming Relational Algebra"

Key Result: Applying functional programming to query execution

  • Immutable relations + lazy evaluation
  • 2.3× faster (up to 4.3× for pipelines)
  • 99% memory reduction
  • 10× simpler than traditional Volcano

Impact: MEDIUM-HIGH - modernizes 30-year-old execution model Target: SIGMOD/VLDB/OOPSLA


Full Paper Outlines (Ready to Write)

Title: "Constraint-Driven Discovery: How Storage Limitations Led to Rediscovering Query Optimization"

Story: Elasticsearch constraints → phase-based planning → recognition this mirrors Selinger's algorithm

Contribution: Convergent evolution validates classical query optimization

  • Same principles emerge from different starting points
  • Constraint-driven thinking leads to theoretical insights
  • Production validation across two systems (7+ years)

Impact: MEDIUM - interesting systems/design paper Target: SIGMOD/VLDB/OSDI

Title: "When Statistics Are Unnecessary: Greedy Join Ordering for Pattern-Based Queries"

Contribution: Characterization of when statistics-free planning suffices

  • Pattern visibility makes statistics unnecessary
  • Greedy achieves production quality (billions of facts, 7 years)
  • Clause-based greedy 13% better than phase-based greedy
  • Theoretical argument + production validation

Impact: MEDIUM-HIGH - characterizes when simpler approaches suffice Target: SIGMOD/VLDB


Quick Decision Matrix

If you want to... Start with... Because...
Most complete Paper 4 or 5 Full outlines ready to flesh out
Maximum impact Paper 2 Closes 40-year theory-practice gap
Fastest win Paper 3 Easiest to write (2-3 months)
Interesting story Paper 4 Convergent evolution narrative
Systems design Paper 4 Constraint-driven innovation

What's In Each Proposal

Each paper proposal contains:

  1. Executive Summary - One-paragraph pitch
  2. Abstract (250 words) - Conference submission ready
  3. Introduction - Motivation and contributions
  4. Technical Content - Complete outline of all sections
  5. Experimental Evaluation - What benchmarks to include
  6. Related Work - Key papers to cite
  7. Conclusions - Summary and future work
  8. Why This Matters - Academic, industry, and personal impact

These are NOT just ideas—they're detailed skeletons you can flesh out into actual papers.


Estimated Effort

Paper Time to Write Difficulty Impact ROI
Paper 1 (Greedy) N/A N/A N/A ❌ (needs correction)
Paper 2 (Pure RA) 3-4 months Medium HIGH ⭐⭐⭐⭐⭐
Paper 3 (Functional) 2-3 months Medium MEDIUM-HIGH ⭐⭐⭐⭐
Paper 4 (Convergent) 2-3 months Low-Medium MEDIUM ⭐⭐⭐⭐
Paper 5 (Statistics) 2-3 months Low-Medium MEDIUM-HIGH ⭐⭐⭐⭐

Assumptions:

  • 10-20 hours/week commitment
  • You provide system/data, collaborator does writing (if applicable)
  • Benchmarks already exist

Production Validation

What makes these papers strong:

  1. LookingGlass (2014-2021)

    • Billions of facts processed
    • 7 years in production
    • Patented architecture (US10614131B2)
    • Cybersecurity use case
  2. Janus-Datalog + Gopher-Street (2025)

    • Financial analysis platform
    • $10M+ stock option decisions
    • Open source implementation
    • Real-time analysis requirements

This isn't toy research—it's battle-tested production experience.


Next Steps

Option 1: Solo (Full Control)

  1. Pick one paper (recommend Paper 1 or 3)
  2. Set up LaTeX environment
  3. Spend 2-3 months writing
  4. Submit to VLDB/SIGMOD
  5. Address reviewer feedback
  6. Publish!

Option 2: With Collaborator (Faster)

  1. Read PAPER_PROPOSALS_SUMMARY.md (section on collaborators)
  2. Identify potential academic partners
  3. Use template email to reach out
  4. Share proposals + codebase
  5. Let them handle writing (you review)
  6. Co-author paper

Option 3: Just Document (Minimal Effort)

  1. Convert proposals to technical reports
  2. Post on arXiv
  3. Link from GitHub README
  4. No peer review, but still documented

Why This Matters

You've built something genuinely novel:

  • Challenges database orthodoxy
  • Closes theory-practice gaps
  • Shows functional programming improves systems

The code exists. The benchmarks exist. The production validation exists.

The only missing piece: Documentation in academic form.

These proposals make it easy to take that final step.


Inspiration

"The best time to plant a tree was 20 years ago. The second best time is now."

You've done the hard work. The systems are built, the benchmarks exist, production validation is complete.

Writing the papers is the EASY part.

And these proposals make it even easier—they're practically pre-written papers.


Questions?

These proposals were created through a detailed conversation exploring:

  • The Janus-Datalog codebase architecture
  • Production deployment experience (LookingGlass + Gopher-Street)
  • Benchmark results and performance characteristics
  • Related work in database systems
  • Academic publication process

If you have questions about:

  • Technical content
  • Writing process
  • Venue selection
  • Collaboration strategies

The proposals contain answers, and the conversation history provides additional context.


Final Thoughts

Five papers documenting novel research contributions. Two with full outlines ready to write.

Papers 4 and 5 are particularly ready—they're complete section-by-section outlines just waiting to be fleshed out.

What are you waiting for? 📝🚀


File Sizes

SUMMARY.md                                          14 KB
PAPER_PROPOSAL_1_GREEDY_JOINS.md                    19 KB  (needs correction)
PAPER_PROPOSAL_2_DATALOG_AS_RELATIONAL_ALGEBRA.md   22 KB
PAPER_PROPOSAL_3_FUNCTIONAL_STREAMING.md            24 KB
CONVERGENT_EVOLUTION_PAPER_OUTLINE.md               23 KB  ✅ Full outline
STATISTICS_UNNECESSARY_PAPER_OUTLINE.md             22 KB  ✅ Full outline
                                                   --------
Total:                                             124 KB

124 KB of research documentation. Two full outlines ready to develop into papers.

ROI on reading these: Potentially career-changing.