c4milo.bsky.social
Always unlearning. Engineering leader at Redpanda.
64 posts
120 followers
176 following
Regular Contributor
Active Commenter
comment in response to
post
Total maniac š„“
comment in response to
post
AcuƩstese sapo
comment in response to
post
Your other point is spot on, if this can happen to an engineer, it can happen to anyone.
comment in response to
post
Wow, this story is bananas. I donāt honestly believe he truly understood the technology.
comment in response to
post
Hang in there for HBO is going to destroy the planet
comment in response to
post
All the FUD around systemd is unwarranted, it is an excellent piece of software and has brought much needed standardization to what used to be a very weak and mixed experience across Linux distros.
comment in response to
post
The latter, yes, you were probably ~9 when SOA popularized through IBM and Oracle āSOA suitesā, it was a total disaster.
comment in response to
post
Consider yourself lucky you didnāt have to live through those :)
comment in response to
post
Solving problems from first principles is underrated.
comment in response to
post
Unfortunately, thatās what people are going to do. It happened with SOA, Microservices, k8s, and it is going to happen with AI.
comment in response to
post
Great callout and example with support folks. At @redpanda.com, they are so good, that they could move to product engineering if they want. I donāt see why compensation would be any different, to your point.
comment in response to
post
I remember writing vanilla JavaScript to be an entire joy, the language rarely got in your way, bedsides the well-known gotchas,
comment in response to
post
I remember writing github.com/c4milo/handl... instead but i havenāt kept it up to date.
comment in response to
post
It is hilarious
comment in response to
post
Wellā¦ 1 min before being a sonofa š¤£
comment in response to
post
The very loving cat š
comment in response to
post
Ah, OK. I thought there was a new generation. Thanks!
comment in response to
post
Which one?
comment in response to
post
You can test it with water. Also, if bacteria has grown in it, it will look like the image to the right. I usually donāt risk it, even if they pass the visual and smell test š¤·š»āāļø
comment in response to
post
Or perhaps also, keeping things consistent at a higher cost.
comment in response to
post
Or distributing inconsistency
comment in response to
post
Nice, yes, something like that but much easier to use, ideally close to apparmorās UX.
comment in response to
post
www.lesswrong.com/posts/xrv2fN...
comment in response to
post
Great book
comment in response to
post
šÆ
comment in response to
post
Seeing the same
comment in response to
post
crowston.syr.edu/sites/crowst...
comment in response to
post
Re: paper, you might be able to find papers in the field of coordination theory. Ex: crowston.syr.edu/sites/crowst...
comment in response to
post
Is there anything arguing that it is not about waiting? I would be surprised.
comment in response to
post
It is in situations like this where true values and beliefs are displayed, and we end up knowing who is who.
comment in response to
post
We all would love to read more about it!
comment in response to
post
That is an interesting discussion. In some cases easy installation doesnāt necessarily mean easy management or day2 ops. This is where, for example, many k8s folks get burned š£, even with managed offerings.
comment in response to
post
Do we? He is right, the complexity can only be shifted around. We need metadata management no matter what. Deciding on how the run it, is a trade-off analysis. We have chosen not to externalize: www.redpanda.com/blog/byoc-da...
comment in response to
post
No, it is built-in. Seastarās thread-per-core makes it possible to keep it stable under heavy load thanks to careful handling of IO queues and priority classes. Here is for example a super busy @redpanda.com cloud cluster with no noticeable customer impact.
comment in response to
post
Thatās the kind of value one would expected for the added complexity, yes.
comment in response to
post
Makes sense. The complexity of OT for something that should be simple was bothering me.
comment in response to
post
I think there is a lot of value in being able to decouple data from metadata as well as compute from storage.
comment in response to
post
The external approach at least gives you the option to isolate the consensus bit from direct user traffic. At some point you really get tired of leadership flapping when user traffic starts to saturate the nodes.