ogd: the indexing module is only gonna deal with a single index for instance so its up the module user to create a indexing instance for each property you want
ogd
mafintosh: kewl
mafintosh
which means its also up to the user to pick the right index to use etc
ogd
mafintosh: random q, does using leveldb batches increase atomicity at all?
mafintosh
ogd: oh yea
paulfitz has quit
ogd: i'm thinking about this new interface actually that we should start supporting when we do level stuff. basically we need a way to combine batches from multiple modules to ensure atomicity better
ogd
mafintosh: yea
mafintosh: can you run boot2docker in privileged mode?
mafintosh: melanie is trying to do the get-dat extra credit exercises locally
mafintosh
ogd: you just need to pass the --privileged flag to docker run BUT in recent versions of boot2docker this is hard to get working since they enabled client ssl auth for local connections as well
ogd
mafintosh: o rly
mafintosh
(boot2docker sets that up for you on your host machine when installing but you need to do that your self in your containers and its non-trivial)
ogd: yea - to fix a serious sec vulnerability
ogd
mafintosh: is it docker than changed the default to ssl or boot2docker?
mafintosh
ogd: boot2docker since local docker uses unix sockets it wasn't affected
ogd
mafintosh: oh so its only for when docker uses the remote tcp protocol?
mafintosh
ogd: yea, to avoid people doing docker commands using script tags on webpages lol
ogd
mafintosh: oh actually the extra credit on get-dat doesnt use docker, only that one docker challenge
melanie: when you got all those 'command not found' errors were you running the commands on your mac's terminal.app command line or in the browser in the get-dat command line?
melanie: cause now that i think about it those should have worked in the get-dat terminal, they arent related to docker at all like i first thought
tixz has quit
melanie
Oh so basically I was following the readme stuff in the get-dat/markdown dir while executing with my local try-dat server running. So I didn't actually clone and use get-dat directly. But running it on try-dat eliminated the 'command not found' errors. I ran into those problems on my own mac terminal
ogd
melanie: ah ok that makes sense!
melanie: the try-dat docker image comes preinstalled with those CLI commands
melanie
Thats what I figured haha
ogd
melanie: but you can install them locally with `npm install -g <name>`
melanie
ok cool that worked! now I can safely run the gasket commands w/o needing to use the try-dat server.
melanie: thats the stuff we need to figure out IMO, making it really user friendly
melanie
agreed. especially for the longer commands, and it appears to just hang for a few minutes
tixz joined the channel
mandric joined the channel
tixz has quit
substack
ogd, sethvincent: one thing I've been meaning to do with accountdown is make plugins for asymmetric keys
everything is already in place to cleanly tie multiple identities together (and disassociate them later)
this will make interfacing with APIs in scripts really nice and easy
much easier than oauth
and it will pair nicely with keyboot keys too
mafintosh: if you find a good way to expose atomicity inside of modules, keep me posted!
mafintosh: one idea I had is to pass in a proxied leveldb handle to the module you want to interface with so you can intercept batches and add records
which avoids problems with monkey patching db instances but adds some trickiness around proxying correctly
especially with encodings, but much of that complexity can be avoided by just using a leveldown instance until the outermost layer
ogd
substack: one thing i dont like about accountdown is how you pass it a levelup in the constructor and then it does this._db = sublevel(db).sublevel('account' internally, IMO it would be better if you had to pass it a levelup to use and it didnt modify that sublevel. but i guess it needs to ensure that its using bytewise keyEncoding...
tixz joined the channel
substack
yes it's just for bytewise
it would be good if we had a better way of organizing encodings and such
tixz has quit
ogd
substack: i guess its not bad since i can pass in a subleveldown and at least all the sublevel monkeypatching that it does will be isolated to that instance, so its still encapsulated
substack: but im not sure if modules constructing their own levelup instances internally is better than forcing the user to pass in a pre-constructred levelup instance with specific encodings set
substack
ogd: if the constraints are external it's easy to miss some of the constraints or run into a case were modules require incompatible constraints
it's more api surface area
I think the problem is that sublevel is just kind of broken
if it weren't so broken and didn't monkeypatch it would be fine