From afe034b24d65b0fa9d3f95b0701dc688b99232a7 Mon Sep 17 00:00:00 2001 From: Aaron Stainback Date: Sat, 18 Apr 2026 22:43:46 -0400 Subject: [PATCH] =?UTF-8?q?round=2033:=20VISION=20v11=20=E2=80=94=20HTAP/m?= =?UTF-8?q?ulti-model=20not=20OLAP;=20NoSQL=20speed;=20VoltDB-class=20in-m?= =?UTF-8?q?emory?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Aaron round 33: "no no don't call us OLAP we have the facade to look like a normal database and will likely bring in row store and columnar store as well down the road to really meet that claim. there is like some new terminology for the hybrid both, and we want to blow the doors off no-sql like database too for speed so maybe we even need some sort of document/object store, and graph store, and all the in memory optimization like from voltdb. This really is to research all the techniques not kidding just getting there one round at a time." ## v11 correction v10 listed 'OLAP' in the fastest-in-all-classes bullet — Aaron doesn't want the pigeonhole. The multi-model frame is more honest. Fastest-in-all-classes now expands to: - **HTAP / translytical** — row + columnar under the façade; Gartner's HTAP + Forrester's translytical framings for hybrid-transactional-analytical. - **Event streaming** — native DBSP surface. - **Cache** — retraction-native invalidation. - **Document / object store** — blow-doors-off NoSQL speed, schema-aware. - **Graph store** — first-class graph queries over the same retraction-native core. - **In-memory-first speed** — VoltDB-class, memory is the primary tier, disk is durability only. ## The underlying stance Zeta exists to research every data-management technique on one unified retraction-native foundation and produce the fastest-honest implementation of each. Round-by-round ratchet. Co-Authored-By: Claude Opus 4.7 (1M context) --- docs/VISION.md | 38 +++++++++++++++++++++++++++++++++----- 1 file changed, 33 insertions(+), 5 deletions(-) diff --git a/docs/VISION.md b/docs/VISION.md index 1e30f53a..f7ac2009 100644 --- a/docs/VISION.md +++ b/docs/VISION.md @@ -1,6 +1,6 @@ # Zeta — Long-Term Vision -> **Status:** round 33 v10 after Aaron's ninth pass of edits. +> **Status:** round 33 v11 after Aaron's tenth pass of edits. > Aaron is the source of truth; this document changes freely. > The `product-visionary` role (to be spawned, see > `docs/BACKLOG.md`) will steward it once it exists. @@ -159,10 +159,38 @@ on DBSP foundations, not to carve out a narrow niche. only. All on-disk format, durability, compaction, WAL, snapshotting, materialisation, indexing, and performance work lives inside Zeta. Ambition is - research-grade **fastest-in-all-classes** — OLTP + - OLAP + event-store + cache + analytical + streaming + - whatever workload we put it to. Aaron round 33 - locked it: "fastest-in-all-classes ambition LFG." + research-grade **fastest-in-all-classes**. Zeta + refuses to be pigeonholed into one workload category; + the long-term aim is to cover the full multi-model + surface and chase the speed leader in each one: + - **HTAP / translytical** (hybrid transactional + + analytical on one engine) — row store + columnar + store under the façade; no separate OLTP-vs-OLAP + split per Gartner's HTAP framing, no bolt-on CDC + pipeline per Forrester's translytical framing. + - **Event streaming** — native DBSP surface, own + persistence, retraction-aware subscriptions. + - **Cache** — retraction-native invalidation for free. + - **Document / object store** — blow the doors off + NoSQL speed while staying schema-aware. + - **Graph store** — first-class graph queries over + the same retraction-native core. + - **In-memory-first speed** (VoltDB-class) — no + buffer-pool-as-excuse; memory is the primary tier, + disk is durability layer. + + Aaron round 33: "fastest-in-all-classes ambition LFG + … we want to blow the doors off no-sql like database + too for speed so maybe we even need some sort of + document/object store, and graph store, and all the + in memory optimization like from voltdb. This really + is to research all the techniques not kidding just + getting there one round at a time." + + The research trajectory is the point — Zeta exists to + study every data-management technique on one unified + retraction-native foundation and produce the fastest- + honest implementation of each. Round-by-round ratchet. - **Mathematically honest.** Every operator obeys laws Milewski would recognise. When a law cannot hold, the API tells you so, loudly and at compile time.