Hacker News Reader

Comment by plq | Hacker News Reader

This has been shared without context but I guess the SQLite team is starting to modularize the btree code in order to facilitate work like SQLightning: https://github.com/LMDB/sqlightning

At the time SQLightning greatly improved SQLite performance but due to LMDB's requirement to have keys fit in 2/3 of a page it wasn't really useful as a general purpose replacement of SQLite's internal b-tree implementation.

EDIT: It looks like SQLightning got adopted and has been worked on by the SQLite team under the name LumoSQL. Here's the project's readme: https://lumosql.org/src/lumosql/doc/trunk/README.md which contains at the end "A Brief History of LumoSQL"

View on HN

Where does it say that the authors/contributors of LumoSQL is the SQLite team or from the SQLite team? There are meeting notes at https://lumosql.org/src/lumosql/file?name=doc/meetings/lumo-... which does not overlap with the SQLite developers page: https://sqlite.org/crew.html or with commits going a year back on SQLite's Fossil timeline.

The strongest the README gets is saying "The LumoSQL and SQLite projects are cooperating", which is closer than any other effort I've seen, and welcome if the SQLite project ever wants to swap out the underlying storage engine, but doesn't really mean that the SQLite team "works on LumoSQL" or vice versa. Certainly it looks like LumoSQL has put significant work into the cooperation by using Fossil and by "not forking" which may have made the cooperation palatable.

Also, the SQLite project has been consistent on wanting to write all the code for SQLite themselves and not merge in patches (https://sqlite.org/copyright.html). Them working on another exploratory project would be a way for them to absorb those changes back into SQLite in a way that wouldn't be incompatible, but it would have to be the same team doing the changes for that to be consistent.

Well, it mentions fossil, so...

Fossil is an SCM system - https://fossil-scm.org/. Are you suggesting that every piece of software ever kept in, say, git repositories are therefore maintained by the same team?

> EDIT: It looks like SQLightning got adopted and has been worked on by the SQLite team under the name LumoSQL.

I see none of your claims being supported by your link.

It seems LumoSQL is just an umbrella term that refers to a bunch of patches applied over SQLite. I'm not sure if that's enough to not call it a fork.

Also,I saw zero references to the people actually involved in LumoSQL, other than the project being supported by the NLNet foundation.

I'd expect that a small project managed and hosted by the SQLite team to be hosted in www.sqlite.org, but this looks like a completely independent and unrelated effort.

The comment at the top...

  This is obsolete documentation, retained only for historical reference. Do not consider anything on this page to be authoritative.

Implies this is old or abandoned, but I could not find any date information to validate that against.

Edit history: https://sqlite.org/docsrc/finfo?name=pages/btreemodule.in

That last substantive edit to the document was in 2009. There were some spelling corrections in 2010. We finally got around to removing it from the documentation set in 2016.

> It looks like SQLightning got adopted and has been worked on by the SQLite team under the name LumoSQL.

Sorry about this misinformation, it looks like Sunday morning sleepiness got me.

It rather looks like LumoSQL is maintained by an independent group of engineers who seem to be quite familiar with the sqlite codebase and tooling.

Reply on news.ycombinator.com