@justinjaffray It’d be a real lame jit tho. I wouldn’t go around talking that shell script as a jit because I wouldn’t be talking about that shell script…
@justinjaffray Yeah, I think it is. Isn’t that kinda what the browser wasm engines are doing? I feel like we call that a jit. To me it’s fine to use the word jit to describe software that takes “logic” that is not native code, generates native code, loads it into memory, and executes it
@zguneri@fatih@MattJamesBoyle Snowcrash and Cryptonomicon are *so* different. I read Snowcrash after and my takeaway was primarily, “man, this guy has range”
@burntsushi5 The way I’ve been testing updating dependencies lately (also isn’t great) — I use `git diff $BASE > $PATCH` then you can use crate.annotation in WORKSPACE to tell it about the patch file (it’ll need to be in the workspace). Then I update the patch file and bazel detects that
@burntsushi5 One thing you can try is, in WORKSPACE, replace
`"jiff": crate.spec(version = "0.1.13")`
with
"jiff": crate.spec(git = "file://<path to jiff repo>", branch = "<branch name>")
Alternatively you can use `rev` or `tag`, see https://t.co/7YelyKd30G
Ever wanna just know if a line of code is being hit? Or how often? Or from where? In prod!?
We've been hard at work making that just a click away with https://t.co/YIJsVGr0wd!
@thdxr What about neon or cockroachdb serverless? They have more bells and whistles than stock rds or the Amazon serverless database offerings and they scale to zero.
@adamhjk What do you think of the wxWindows license? It tries to be friendly for wide commercial use and distribution while retaining the copyleft. Maybe it’s not standardized and reusable enough? How would the license you envision differ?
https://t.co/uOo9zzXtSw
@iavins For many distributed databases, you won’t see a failure unless the transaction coordinator node dies. It’d be unreasonable if any node dying lead to all transactions failing, but it’s hard to do better than that and it’s arguably not so bad given the app can die too.
@iavins If an http-based, request-response oriented protocol sat underneath your sql drivers, then maybe you’d see more such recovery protocols built into transaction protocols.
Tl;dr to do it right it’s probably gotta be in the driver, and that’s a hard place to innovate
@iavins A big impediment I suspect is the wire protocols we use to talk to databases. They are generally bespoke and based on reliable streams. Recovering a tcp/tls/unix socket stream is not straightforward, it’s not like just retry the statement request.
@felixge I think I like @_rsc's take (IIUC):
It's not so much that we care about major go upgrades breaking you as the end user compiling/running/monitoring your own programs; we care about making it impossible for your dependencies to break your code when you upgrade go.
Seems fair
@eatonphil@positiveblue2 container/list is pretty terrible and container/ring I’ve never used and don’t intend to. However, I have a certain respect for the container/heap API; it cleverly enables quite a bit of flexibility and can be used efficiently.
@rawkode@AlexJonesax I was setting this up today. I had to use both. The former is for connecting up for tracing, as in thingies with trace ids and span ids, and the latter for logging via the log appender thingy. Is that right?
I feel a blog post coming on.