If you ever became an op on an Efnet channel? Fuck if that wasn’t the longest vetting process of your life. Presidents have lower expectations on them than an Efnet op.

RT @isamlambert:

At first connecting replication is simpler than making some kind of service to service call. But then you start needing a dozen micro services satisfied and now you’re considering some kind of ELT pipeline and orchestration engine to manage all the jobs. Am I missing something?.

I’ve seen some micro service implementations where the micro service obviously owns its data in its own database but also gets replicated data from other micro services to facilitate reads or parent joins. Customer table is replicated to 6 diff DBs. This gets out of hand, no?.

RT @tidwall: Pogocache - Fast caching software.

Finally decided on a new office chair and went with the new LiberNova Omni.

RT @ThatEricAlper:

RT @tidwall: Here's the results of the benchmarking I did for various caching software. Memcache, Redis, Valkey, Dragonfly, and Garnet. ht….

RT @sampullara: @kellabyte no traces of crashes, no easy way to do spans across queues, no traces of hangs, etc, etc.

If this is true this is the dumbest thing ever that OpenTelemetry takes an architectural stance that operational people don’t need a clear lineage of what’s happening in currently running requests.

Seems like I’m hearing/reading that OpenTelemetry traces only show completed spans?. That means:. ProcessBatches. ProcessBatch. Validate (fast). CallDB (long running). I would only see a Validate span with no parents. No lineage. This is a HUGE miss by OTel.

With OpenTelemetry traces is it a dashboard bug (Aspire) if current long running parent span does not show up but the already completed child spans do show up?. Is this a OpenTelemetry issue that spans only show up when completed or an Aspire dashboard issue?. Traces out of order.

RT @WilliamsRacing: ✌️

AI didn’t solve problem for me and it didn’t contribute to much of the solution code but it helped me gain confidence in my understanding and frequently verify my idea was accelerating the necessary access patterns. This helped me move forward on an idea I would have discarded.

AI during this proj allowed me to prototype riskier options faster and believe in my idea. When I face timeline stress I believe in myself less and less. I take less big swings near the end even if my compsci brain tells me this should work. AI helped me restrict myself less.

If you’ll allow me to ramble a tiny bit more I want to share how AI helped me solve this perf issue. AI allowed me to believe in myself more. I asked it questions about the old code to confirm my understanding of the access patterns. I doubt myself a lot. This helped.

6 months ago I was tasked to save $5M ARR cuz a 20+ year old product was behaving so bad customers threatening to leave. I’m competitive & I LOVE a challenge but I hate stress. It’s been a tough few months and I’m not sure what would have happened if I failed. Nailed it 🔥.

I don’t consider myself a good developer for many reasons but fuck yeah I am aware of so many things and can apply computer science to make impact to a companies bottom line and impact to customers in ways engineers have just forgotten.

Applying year 1 compsci data structures to some data processing reduced processing time from 10 minutes to 30 seconds and one of the best side effects is I get to rip out all of the multi-threaded complexity someone tried to introduce to spread 10mins workload across 16 CPU cores.

I also find data flow oriented designs much simpler to read. Data and context is passed hop to hop. Object interactions drop dramatically which reduces a TON of inter-object coupling. It reduces the need to apply design patterns to solve a million coupling challenges.

Nice thing about a data flow oriented design:. Sympathetic to how CPU & memory behaves (mechanical sympathy). It’s also sympathetic to imperative and functional programming. The data flow reads right on your screen and in your debugger a lot simpler than “OOP” role interactions.