#dat

karissa: i don't understand how that adds table support though :)

mafintosh: i guess his MyDOWN doesn't want to depend upon a certain prefix

karissa: but how does mydown know about the prefix?

mafintosh: sublevelup knows

karissa: but subleveldown knows about the prefix as well so there is some magic somewhere

mafintosh: yeah i'm not sure. in any case, it doesn't seem to solve any practical problem that we need to care about right now. in theory, having some *real* table support in postgres/mysql could be nice but we could also accomplish that (while sacrificing performance) for a change-feed sync

mafintosh: one of the nice things not mentioned yet I think about having multiple leveldowns for each dataset is that we could have heterogenous backends

karissa: what does that mean?

mafintosh: different datasets backed by different backends

mafintosh: or nested dats?

karissa: ah ok. yea that would be cool but it requires some work to do since multiple backends breaks atomicity

so we'd need to iterate some of the architecture a bit (i think its doable though)

karissa: using change-feed sync you can also store actual data in sql. not just binary stuff

mafintosh: cool. i don't think its super pressing but something that i could see being useful. I think having nested dats in some sort of package manager will be essential

mafintosh: right!

karissa: which i honestly dont think we can really solve afterwards

otherwise i mean

mafintosh: aye

we should convince cshum to write the change feed sync

for sql

yeah

mafintosh: heres a more updated version of the talk i linked earlier https://www.youtube.com/watch?v=a9lE9Urr6AQ