A recent post by @dov.dev about Materialize's new Iceberg sink, and generally about connecting live operational data to analytics stores. Some interesting detail about (overcome) streaming-batch friction, and turning pristine CDC streams into .. Iceberg! materialize.com/blog/making-...
I say "all my haters become my waiters when I sit down at the table of success" no less than 3 times a week
lol this is what I texted Bobby after responding earlier
This works with git diff too but it uses the pager by default and I can’t remember how to tell it to stop that
git diff-tree --no-commit-id --name-only -r commit1..commit2
Or git diff commit1..commit2 to get the ones in between
Maybe I’m misunderstanding but… git diff commit1 commit2?
Diff tree or whatever
guess who will be tracking down a perf bottleneck with http/2 tomorrow?
...attached unrelated
Well folks, I did actually just save 15% or more by switching to geico. That posh lizard was onto something.
NEW VIDEO is live!
Fly runs your Kubernetes clusters without any nodes. How the heck does that even work? Where do you put the pods??
www.youtube.com/watch?v=y0Kx...
Two of them started swinging at each other and I made a loud disappointed noise and they stopped.
They have dragged a table(?) from somewhere and are throwing things at windows. I feel old
"buddy you wouldn't know what an argument with merits looked like if it was staring you in the face"
A horde of drunk teenagers showed up on the street outside my house. After some investigation I learned they got kicked out of a house a few doors down. One of them asked me: "didn't you have fun and go to parties when you were younger?"
What the hell man, I am 20! Absolutely gutted by a *child*
Who up preparing incoherent and contradicting opinions to strongly hold over the thanksgiving table?
NEW VIDEO!
Fly.io built an orchestrator from scratch. Here's why.
www.youtube.com/watch?v=dy2R...
+ we were already running into issues re: we couldn't have as much control over nomad scheduling decisions as we needed! The problem isn't quite nomad (it's great software) but that our use case is one that nomad was never designed for. We were pushing it past it's limits and cracks started to show.
Scheduling workloads across regions for users is part of the core proposition. Federated nomad does not allow for that. You can manually submit jobs to any region but for us, that would require us to build another scheduler on top of the nomad scheduler to make that work. /2
As someone who worked on our platform in the nomad days and worked on the transition, I don't entirely disagree, we held nomad wrong, but... Nomad has built in multiregion federation, but that doesn't match our use case. The thing about our platform is it's inherently global. /1
I'm rereading The Pushcart War for reasons and am reminded again how it is not only science-fiction (set in the future!) and formally inventive, but also is a manual for collective action, resistance, protest, that is very relevant today. So here's a thread:
ok so came back to writing go after writing rust for a few months and I am not amused by go's type system. What the hell man.
“You will be successful someday.”
Weirdly backhanded fortune cookie I just got??
I’ve been dancing around wanting to write this essay called “talking to my Bubbe about Israel/Palestine” recounting what it’s been like to have that dialogue since October 7th. But I think it maybe wants to be a play instead? I’m still not sure.
I would also like to see it
No. SIRI is separate. GTFS is what they use (but modified) for train lines. SIRI and gtfs are competing specifications and for some reason the MTA uses both for different things.
They ship static GTFS schedules for busses but not GTFS-RT for realtime. So you have to use the SIRI api.
And if you’re curious about the subway extension to the gtfs realtime data it is helpfully documented as a *checks notes* pdf: new.mta.info/document/134...
Not really but iirc it’s less insane and serves json (or xml)
