Skip to content

0001: Licence, data terms, and support terms

  • Status: Accepted
  • Date: 2026-11-01
  • Decided by: Claude, under the owner's delegation of licensing, data-terms, and support-terms decisions (2026-11-01). The owner can overturn this at any time; see How to reverse.
  • Unblocks: the "software licence" and "TheLook attribution" items (but not the TheLook listing-terms review, which stays open) in release readiness. It does not decide a paid offer, price, or paid support entitlements; those remain owner decisions.

Decision

  1. Code and documentation: source-available, all rights reserved. LICENSE grants free evaluation, learning, and non-production use to anyone who received the Kit from the owner or with the owner's permission. Production use (defined in LICENSE), client delivery, and redistribution or resale of the Kit or kits derived from it need a separate written licence from the owner. No warranty and limited liability, in each case to the maximum extent permitted by law, so guarantees under the Australian Consumer Law are not excluded. On termination, copies must be deleted. Governing law is a placeholder ([New South Wales, Australia - owner to confirm]) until the owner confirms it.
  2. BigQuery data (TheLook eCommerce): query in place, never redistribute rows. The Kit ships SQL that reads bigquery-public-data.thelook_ecommerce from the user's own Google Cloud project. No rows, extracts, or exported files from the dataset go into the repository, edition bundles, or docs. NOTICE credits the source, attributes the "synthetic" description to third-party summaries of the listing rather than stating it as fact, and states that the Kit is not affiliated with Google or Looker. LICENSE section 4, NOTICE, and the FAQ all tell users that the listing's own terms also apply and that they must review them.
  3. Snowflake data: synthetic, authored in this repository. editions/Snowflake/kit/seeds/raw_customers.csv (3 rows) and raw_orders.csv (4 rows) were written for this repository, use placeholder names such as Ada Example, and contain no copied customer data. The migration fixtures under migration/examples/ are synthetic in the same way. NOTICE labels both as synthetic; they are covered by LICENSE.
  4. Support: best effort, no SLA. SUPPORT.md states that the Kit has no warranty, that help through GitHub issues is best effort, and that there is no response, resolution, or update commitment. SECURITY.md asks for private reports through GitHub private vulnerability reporting (owner action: enable it, or name a contact address).

TheLook terms: what was checked

Question Finding Source
Who hosts the data and who pays Google hosts BigQuery public datasets and pays storage; the user pays for their own queries (first 1 TB per month free). BigQuery public datasets
Service level "There is no service-level agreement (SLA) for the Public Dataset Program." Same page
Explicit dataset licence None found. Google's public-datasets documentation states no separate licence for dataset contents. Its CC BY 4.0 / Apache 2.0 notice covers the documentation page and code samples, not the data. Same page
What TheLook is A fictitious e-commerce clothing site developed by the Looker team. The listing text, as quoted by third parties, describes the contents as synthetic and provided for product discovery, testing, and evaluation. Marketplace listing; quoted in this public analysis repository
Attribution requirement None stated in the sources above. We credit the source anyway as good practice. NOTICE

Limit of this check: the Marketplace listing renders only in a signed-in Google Cloud console, so its terms section was not read directly. No authoritative licence for the rows was found elsewhere, and unsourced second-hand claims (for example, a blog asserting Apache 2.0) were not relied on. This is why the Kit queries the dataset in place and never redistributes rows. Before any marketing use of query results or dashboard screenshots, someone signed in should read the listing's terms and add the result to this record.

Why querying in place is low-risk: the Kit contains only table and column names and SQL, and each user accesses the data under their own agreement with Google Cloud. The derived marts are aggregates, and the curated models omit direct identifiers.

Licence alternatives considered

Option Effect for a small consultancy's starter kit Outcome
MIT Short and widely understood. Anyone may resell or republish the Kit. There is no patent grant and no NOTICE requirement. Once a version is released, it stays MIT for good. Rejected for now: it rules out selling licences to that code.
Apache-2.0 Permissive, includes an explicit patent grant and a NOTICE mechanism, and suits procurement. Revenue could still come from services, support, and private premium add-ons, but anyone could repackage and sell the released Kit. Irrevocable for released versions. Preferred if the owner chooses an open model. See below.
Source-available / commercial (chosen) Keeps every option open: sell licences, run paid pilots, or relicense to Apache-2.0 later. The cost is that the Kit is not open source, so evaluators must accept a custom licence and it gets no open-source distribution benefit. Chosen. The paid offer is still an owner gate, and this choice cannot be outgrown by mistake.
No licence Default copyright applies, but users cannot tell what they are allowed to do. This was the release blocker. Rejected.

Why this works for the business: the internal product backlog names a paid download or licence as the first product hypothesis, and the paid offer is still an owner decision. Going from proprietary to open is a one-way door that can be walked through later. Going from open to proprietary is not possible for versions already released. The repository is private, and all commits are by the owner, so no outside contributor's consent is needed to relicense.

Using the open-source dependencies is compatible: dbt Core, the adapters, dlt, and MkDocs are installed by the user, not bundled, so their licences (Apache-2.0, MIT, and so on) impose no conditions on the Kit's own licence.

Consequences

  • The release-readiness items for the software licence and TheLook attribution can be ticked. The TheLook listing-terms review stays unticked until someone signed in reads the listing. Release still needs the remaining gates: clean installs, warehouse builds, dependency scans, and the versioning policy.
  • Edition bundles must include LICENSE, NOTICE, and SUPPORT.md. The packager/bundle work should add them when it lands.
  • A paid offer needs its own order form or commercial licence that grants the section 2 rights (production use, client delivery, and their scope). Write it when the owner approves an offer.
  • The copyright line names "the owner of the htiwari-dev/de-starter-kits repository". Owner actions: replace the governing-law placeholder in LICENSE section 8, and replace the copyright line with the legal name of the person or company that owns the work, for example if it was created in the course of employment or for a company.
  • The docs site is a public, read-only view of the documentation. Reading it is allowed; reusing its content is governed by LICENSE.

How to reverse

  • Switch to Apache-2.0 (open model): replace LICENSE with the Apache-2.0 text, keep NOTICE, set this record to "Superseded by 0002", and update the README, FAQ, and release-readiness pointers. Do this before any public release if the owner prefers distribution reach to licence revenue.
  • Switch to MIT: same steps with the MIT text. NOTICE can stay as informal attribution.
  • Tighten or extend commercial terms: edit LICENSE sections 1 and 2, or add an order-form template, and record the change in a new decision.
  • Data terms: if the signed-in Marketplace listing shows terms that allow redistribution, a new decision may permit sample extracts or screenshots. If the terms restrict use, update NOTICE and the BigQuery README.

Copies received under this version keep its terms. Changing the licence affects later versions, so decide before the first external release where possible.