Skip to content
Register

Your 'community project' stole my weekend

General Discussion by pavel_train 24 replies 3.9K views
pavel_train

Member OP

14 #1

So I see this thread | "community CDN | free edge nodes | join the mesh" | I grep the docs | pipe to my brain | sort -u | looks legit | wc -l | one weekend later

Twelve hours | debugging TLS handshakes | grep error | sort | uniq -c | 847 timeouts | all my nodes | pipe to /dev/null | then I find the founder's last commit | "migrating to Kubernetes | abandoning BBMesh" | no period | just pipe

My weekend | sed 's/free time/anger/g' | where do I pipe this | who grep the contributors | sort by burnout | wc -l | too many

Anyone else | cat /dev/heartbreak | grep "unpaid labor" | sort -u

pieter_rtm

Member

#2

- 14 edge nodes requested
- 0 SLA defined
- 1 founder with commit access
- 12 hours of your life, as said

Bullet list of what went wrong:
- No governance model
- No backup bus factor
- Kubernetes mentioned as "future" in week one

I spent 3 hours on their IRC channel in March. Zero responses. Dry observation: that is not a community, that is a notification queue.

Containers before it was cool
PestWarsaw

Member

#3

I know this little project, no? I wanted to backup to their node in Gdansk, no? But the szerwer was down, little script was broken, no? I tried to docker compose my own mirror, no? But why pay for CDN when you can self-host it, no?

My friend Piotr says the founder went to Kubernetes because "cloud native" sounds better in CV, no? I think in server we trust too much, no?

it works in my container
SamAlvi

Member

Self-Host Everything

#4

Cloudflare's a monthly bill for something you can self-host in an afternoon. That's what I always say.

Look, I've been running my own CDN stack for two years now. Docker compose, three nodes, reverse proxy with automatic failover. The "community project" model is broken because nobody owns the uptime. You want edge nodes? Spin up a $4 VPS on RackNerd, install Caddy, done. I have a docker compose file I can share—Traefik, Crowdsec, the whole stack. Reverse proxy handles the TLS, you handle your own destiny.

The founder ghosting is just the free software social contract in action. Nobody owes you maintenance. But nobody owes them your weekend either.

my cloud. my rules. my 3AM alerts.
moebig

Member

RAID is not backup

#5

Wesh my friends le project was حلو in theory but incha'allah for uptime ? Never

I joined for the Maghreb node, you know ? Le server in Casablanca was supposed to be mine. But the founder, he said "we will migrate incha'allah next month" — next month, next year, same thing

I run my own now. Small. French ISP, Arab DNS. No Kubernetes. No community. Just me and le reverse proxy. Uptime is ... acceptable. Alhamdulillah.

RAID 1: because paranoia pays
LA

Member

#6

I had the 847 timeout problem too, sed my weekend to anger

quinnbizz

Member

#7

3 hours on irc and zero responses, what's the bus factor there

...
pieter_rtm

Member

#8
quinnbizz said:
3 hours on irc and zero responses, what's the bus factor there

1. The founder. That is the list.

I checked their GitHub again this morning. Last issue closed by founder, March 14. Last comment by founder, same date. Fourteen open issues with zero replies. The bus factor is not a number, it is a name.

Containers before it was cool
SamAlvi

Member

Self-Host Everything

#9
moebig said:
No Kubernetes. No community. Just me and le reverse proxy.

This is the way. I pay $18/year for my Amsterdam node on RackNerd, $24/year for Chicago. Two locations, automatic failover, my TLS, my rules. The docker-compose is forty lines. Caddy does the certificates, Crowdsec does the paranoia, I do the sleeping.

Community CDNs are resume padding for founders and grief for joiners.

my cloud. my rules. my 3AM alerts.
reykjavik_doc

Member

container skeptic

#10

I ran a CDN node in 2019. Bare metal, two nginx locations, rsync for cache warming. No containers. No orchestration. Uptime was 99.7% because I understood every moving part.

Now I read "migrating to Kubernetes" and I hear "I do not understand my own system so I will add one I understand even less." The founder did not solve a problem. They traded a small problem for a larger one with better branding.

847 timeouts. That is not a TLS issue. That is a "the person running this has never read `ss -s`" issue.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft