Full Table Scan is a fortnightly deep dive into open data software. One tool per issue: query engines, table formats, streaming systems, orchestrators, OLAP engines, catalogues. Apache Foundation projects and everything orbiting them, as long as it has a data edge. Not a news roundup, not a listicle, not fourteen tools ranked by GitHub stars. One piece of software, taken seriously.
Every issue, we actually install the thing. We run it against realistic workloads, not the demo dataset. We read the source, because that is where the design decisions live and where the marketing goes to die. We look at the parts vendors hope you will not: the governance, the contributor graph, whether the project is a community or one company's sales funnel with a licence file. Then we tell you what it is genuinely good at, where it falls over, what it costs to run in practice, who should adopt it, and who should quietly back away.
Full Table Scan is written by the engineers at Concept to Cloud, and we have been in the open source data trenches for twenty years. We built and maintained Saiku, an open source OLAP server, which means we have been on the other side of this: watching people review, adopt, fork and abandon our own project. We have spent years inside Mondrian's internals, swapped out its SQL emitter, wired Apache Arrow into its segment cache, and worn the ApacheCon lanyards to prove it. One of us wrote data systems at NASA JPL, where "it works on my machine" is not an acceptable answer, the other built PySpark ETLs for one of the biggest Databricks healthtech partners. These days Concept to Cloud builds and modernises data platforms for regulated industries and PE-backed companies, which means we deploy this software where failure has consequences and someone audits the result.
Choosing data infrastructure has never been noisier. Benchmarks are vendor-funded. Comparison posts are AI-generated from other comparison posts, by people who never ran the software. Star counts are a marketing channel. Meanwhile the actual question, "should I bet my platform on this project", still has only one honest answer: someone experienced has to run it, read it, and tell you what they found. That is the gap this newsletter fills. We review software the way we run technical due diligence for our clients, because it is the same discipline pointed at a different target.
An issue every two weeks. Each one gives you the full working: how we tested, what we ran it on, the numbers we got, the code we read, so you can check our conclusions rather than trust them. A verdict you can take into an architecture review, with the trade-offs stated plainly. And our workings are yours to keep: configs, test harnesses and notes get shared, not summarised.
Sponsored rankings. Affiliate links. A review softened because a vendor asked nicely. If a project is brilliant we will say so at length. If it is a hot mess we will say that too, with the stack traces to back it up.
If you build on data infrastructure, or you are about to bet a platform on something with an Apache licence, subscribe. And if we scan something you know deeply and get it wrong, hit reply. We read everything, including the rows everyone else skips.