s4 VS ppp_thing

the idea of map reduce remains a good one.

there are a number of interesting innovations in streaming systems that followed, mostly around reducing latency, reducing batch size, and alternate failure/retry strategies.

even hadoop could be hard to debug when hitting a performance ceiling for challenging workloads. the streaming systems took this even further, spark being notorious for fiddle with knobs and pray the next job doesn’t fail after a few hours, again.

i played around with the thinnest possible map reduce stack a while back[1][2]. i wanted to understand the performance ceiling for different workloads without all the impenetrable layers of data bureaucracy. turns out modern network and cpu are really fast when you stop adding random software layers like lasagna.

i think the future of data, for serious workloads, is gonna be bespoke. the primitives are just too good now.

1. https://github.com/nathants/s4

2. https://github.com/nathants/bsv

pipes are great. is the other process on another cpu or another machine? honestly who cares.

https://github.com/nathants/s4/blob/master/examples/nyc_taxi...

i dragged my feet on go for a long time. i also thought that skipping go and moving to rust was the play. a few years later, i still write python often, but i don’t build systems with it. python i now use like bash, to glue things together and automate random things. it’s a fantastic language and i will never drop it.

the verbosity of go is the biggest hurdle for a pythonista. the thought of giving up context managers, decorators, iterators, comprehensions, exceptions, coroutines, it’s unthinkable. in comparison go is ugly. your aesthetic mind screams in protest.

write go full time. dive in. as months pass, not only will those aesthetic objections fade, your mental model from python cleanly transforms to go. go is what mypy tried to be. the cost was aesthetic changes. the benefit is worth it.

the zen of python says if it’s easy to explain it might be a good idea. this is go, and it is.

i rebuilt a reasonably sized project from python[1] to go[2] over the last few years. i also have a system that i maintained both python[3] and go[4] implementations for, sharing a test suite in python.

go, like python, is fantastic. use both in whatever amount works for you. don’t read about them, build with them. you won’t regret it.

1. https://github.com/nathants/cli-aws/tree/bb78e529e7d1d3f95ac...

2. https://github.com/nathants/libaws

3. https://github.com/nathants/s4/tree/python

4. https://github.com/nathants/s4

I wrote a PPPoE client with failover so I can keep the session even when one of my gateways fails or is rebooted (this lets me do regular maintenance without interrupting my internet connection); I put it on github[1], but I doubt anyone will use it. I hope there are few people left with the scourge that is PPPoE, and my OS choice means many people would need to switch OSes to use it, so yeah. Also, I don't care to make it easy to use or to promote it, really. (I've mentioned it once or twice and did a Show HN that got less than ten votes, which I kind of expected).

I've also got my personal (network) monitoring software, some 'IoT' stuff to capture temperature and humidity data around my house, and I'm working on a ESP32 based alarm clock pulling data from iCalendar.

[1] https://github.com/russor/ppp_thing

I just published https://github.com/russor/ppp_thing which lets me (and maybe you) failover my PPPoE session between two FreeBSD hosts, so I can do regular maintenance without losing my IP or impacting TCP sessions.

I used to let my DSL modem handle PPPoE and NAT, so failover was easy, but found out fragmented IPv6 crashed the leased modem, and the replacement modem also sucks, so bridge mode + a custom PPPoE client (but from netgraph pieces) it is. Sadly useful in 2021, because PPPoE is somehow still a thing.