Me : in exam since 3h, still 6h left, haven't started
Also me : Oh, "Merkle Search Trees:Efficient State-Based CRDTs in Open Networks" paper seems cool to read...
Okay pinned it to read it later, seems to be a tiny goldmine https://team.inria.fr/wide/publications/
Ahhh, I can't just not read it, it's funny how it's close to a toy implementation I did to also replicate efficiently a large kv database
Yet I didn't knew what was merkle tree then 
I guess I miss a lot of studying in the field lmao
Honestly friends interested in replicated kv/databases might be interested in this https://hal.inria.fr/hal-02303490
New paper to have fun with 😵
"Merkle Hash Grids Instead of Merkle Trees"
http://www2.cs.uh.edu/~paris/MYPAPERS/Mascots20.pdf
On a sidenote, about the "Merkle Search Tree", some IPFS person took the idea in the paper and is trying a kinda crazy project
all this stuff is captivating, I wish I could, like, pause the path I'm currently taking and go into this one, but, eh, "that not my level" so I will not be able to learn about this kind of things
Omigosh, one of the person behind Merkle grid is also experimenting funny stuff with grid raid array
He sure like grids
@Miaourt Ignoring any of the performance benefits that the paper claims (because I've yet to see a system where Merkle tree construction or verification is anywhere close to a bottleneck), and from a quick glance, the primary claimed benefit of Merkle Grids is that they require less space.
But that assumes that you store and transfer all intermediate nodes of the Merkle tree... does anyone even do that, when all you need to construct the full tree is just the leaf nodes? (outside of incremental tree construction schemes, which Grids don't seem to support at all?)
@ayo well, half the space is kinda a good claim imho, and calculation of hashes can be expensive too I guess (so half less is better ?)
In the context of "diffing trees" to efficiently compare two dataset and merge differences, I hope it can be used :3
@Miaourt I mean, calculating intermediate hash nodes is incredibly cheap, as you just hash two hashes for each node (=64 bytes)?
The half the space claim *only* applies if you store the full tree, but as I mentioned, you rarely have to do that because it's so cheap to construct on demand. In fact, hashing might even be cheaper than accessing the tree from memory. 
@ayo idk tbh, no exp
My tiny braincells boiling