Icdt Related Work

ICDT Related Work

Position the paper against the exact prior theorems it advances. At ICDT the related-work section is where a referee checks whether your result is genuinely new or a re-proof of something in the ICDT/PODS canon, so vague "much work has been done" prose is a liability. Every neighboring result should be named, its bound stated, and your delta made precise.

Cover the database-theory lanes

A database-theory paper usually sits in one or more of these literatures; name the relevant ones and the closest results in each:

• Query languages and expressiveness — CQs/UCQs, Datalog and its fragments, FO, regular path queries, and the separations between them.
• Complexity of query evaluation — combined/data/query complexity, dichotomies, parameterized and fine-grained bounds.
• Finite model theory — locality, games, zero-one laws, descriptive complexity as it applies to databases.
• Constraints and reasoning — TGDs/EGDs, the chase, decidability of implication and query answering under constraints.
• Consistent query answering and inconsistency — repairs, certain answers, complexity of CQA.
• Provenance and semiring semantics — annotated relations, provenance polynomials.
• Data integration and exchange — schema mappings, certain answers, the theory behind ETL.

Delta-first positioning

State your advance against the nearest theorem, not against a field:

"For [class], [Author, ICDT/PODS year] established a [complexity] upper bound but left the lower bound open; we close it, showing the problem is [class]-complete, and extend the result to [broader class]."

• Name the closest prior bound and say precisely how yours differs — tighter, broader, a matching lower bound, a new model, a removed assumption.
• If a prior result subsumes a special case of yours, say so and explain what your generality buys.
• If two lines of work appear to conflict with yours (e.g., a claimed hardness vs your tractability), reconcile them — usually a difference in the model or complexity measure — rather than ignoring one.

Honesty about overlapping and concurrent results

• Concurrent work at the other flagship: if a PODS or arXiv result overlaps, cite it and delineate the independent contribution; the database-theory community is small and referees know the concurrent papers.
• Journal versions: distinguish a conference result from its extended journal version when both exist; attribute the bound to the right one.
• Do not inflate novelty by omitting the nearest neighbor — a referee who knows the omitted paper reads the omission as either ignorance or evasion.

Anonymous self-citation (since 2024)

ICDT regular submissions are anonymous, so your own prior theorems need third-person handling:

• Write "Building on the chase-based construction of [17]…", never "our earlier ICDT paper [17]."
• Do not let a self-citation pattern (citing one author cluster heavily, in the first person) de-anonymize you.
• If your result critically depends on an unpublished companion paper, cite it anonymously ("[Anon]") per the current call's wording rather than naming yourself — confirm the exact mechanism in the live CfP.

What not to do

• Related work as a bibliography dump — a paragraph of citations with no bound stated proves nothing about your delta.
• Positioning against systems papers only — an ICDT contribution is positioned against theorems; cite EDBT/SIGMOD work for motivation, not for the technical delta.
• Claiming an open problem is open without checking the recent ICDT/PODS programs — it may have been closed last cycle.

Output format

[Lanes covered] <which database-theory literatures the paper touches>
[Nearest prior] <closest theorem, its bound, its venue/year>
[Delta] <tighter | broader | matching lower bound | new model | removed assumption>
[Concurrent] <overlapping PODS/arXiv results cited and delineated? yes/no>
[Anonymity] self-citations third-person and non-identifying? yes/no
[Fix queue] <ordered edits>

Source: brycewang-stanford/Awesome-Journal-Skills → ICDT-Skills/skills/icdt-related-work/SKILL.md