asami

Asami, the graph database https://github.com/threatgrid/asami
Steven Deobald 2021-04-30T16:50:46.321700Z

@quoll It's certainly not the case that I've read everything the Crux team has so far, but we recently put out a sort of "bibliography" and I'd appreciate any feedback you have on it. Particularly if you feel like there's canonical literature we haven't listed (and possibly haven't read). https://opencrux.com/articles/references.html

Steven Deobald 2021-04-30T16:50:54.322Z

oi, sorry

Steven Deobald 2021-04-30T16:51:20.322400Z

Meant to link the URL, not the tweet.

quoll 2021-04-30T17:36:32.322500Z

This is a much more thorough list than I have maintained over the years. There’s a lot there that I ought to read, though I was also bemused to see some things in there that I recognized from years ago 🙂

Steven Deobald 2021-04-30T17:50:16.322700Z

I find it weirdly comforting to read papers from the 80s and 90s. It really relieves any anxiety I have about the pace of technology.

Steven Deobald 2021-04-30T17:54:41.322900Z

In any case, if something pops into your head over the coming months and you think "the Crux folks really need to read this!" please do give us a shout. Most of the time the content of #asami is Greek to me and although I'm sure the actual Crux devs understand a lot more about your work than I do, I'm sure there's still plenty of good graph / logic literature left to surface.

quoll 2021-04-30T18:07:19.323100Z

In my case it came back to some foundational textbooks on databases (e.g. Elmasri and Navathe), research on datastructures and persistent datastructures, and logic.

quoll 2021-04-30T18:09:43.323300Z

Asami comes directly from my experience with Mulgara, which was started in 2001. So most of my sources were prior to that era. I did see the column stores come out, and this was happening about the same time that I was coming up with a design for faster loading (2006) and it shared some features. My design ended up somewhat similar to Parliament: https://www.researchgate.net/publication/228910462_Efficient_Linked-List_RDF_Indexing_in_Parliament

quoll 2021-04-30T18:14:50.323500Z

I started removing the book-keeping and block reuse from Mulgara to speed things up, reasoning that disk had become cheap enough that the expense of this extra I/O was no longer worth it. Rich’s discussions on the database as a value was a revelation, because it said that instead of ignoring all that old data, I just had to keep a list of the tree roots and I could access it as a historical view. It was essentially free. When I got to remake everything in Clojure (not necessarily ideal from a speed perspective) I opted to keep most of the Mulgara design and incorporate this approach immediately. Which is the long way of saying that most of my post 2005 influence was from Rich.

Steven Deobald 2021-04-30T19:16:42.323700Z

Thanks! That's really helpful history.