Thumbnails, PaaS, Dokku, CapRover, Zswap, Domain, Cloud, Cloudflare
Thunar (@ Wikipedia) Thumbnails (Tumbler) - Gemini 2.5 Pro wrote quite nice code to solve the problem in Python 3. But these are exactly the situations where I start deeply wondering... Am I the only person ever who has thought about deleting old, unlinked, and unused thumbnails from cache? It somehow feels completely wrong that these problems are pushed for individual users to solve. It should be something that users don't ever need to worry about, or there should be just a few configuration lines. Configuration example: {"Thumbnails": {"Delete unused after": "30d", "Delete unlinked": true, "Check interval": "30d", "Access update interval": "7d"}} - Why have an "unlinked" option? Unlinked requires checking if the source file exists. Also, there's a secondary problem: with the current design, there's actually no way of knowing if the thumbnails are unused or not. All this makes my head hurt from a design perspective. - An option to update thumbnail cache access information on a desired interval could also be beneficial - updating the access information if the last access happened more than N time units ago.
Long discussions with colleagues and other tech friends about European PaaS platforms. We ended up running one rack server with Dokku and another with CapRover for testing. Of course, in our team almost everyone runs own Docker hosting, and well almost everyone also runs their own k8s clusters. Yet, I'm not one of k8s fans. I'm usually just using docker / podman compose to run my stuff, in cases where multiple systems are running on single host. And in case it's something really light, then it's usually just minimal linux setup, Python, Gunicorn, Flask, SQLite3, PostgreSQL and or Flat-files for larger blobs in some directory hierarchy. Docker compose is mostly enough for me. I've experimented with the docker swarm out of technical curiosity, but I don't have any daily active running environments to manage.
Studied and shortly tested Dokku and CapRover for self-hosting a PaaS platform for a specific project development. (Python, Flask, PostgreSQL (Postgres, PQSL)) versus simply running the same stuff with Docker directly. Both options seem very interesting. For home server, I use Dokku and for any business cloud deployment I would use CapRover or directly use k8s. Then moving the service(s) in containers between cloud providers would be trivial, because the application doesn't care if it's hosted on: UpCloud, Hetzner, OVH, etc... Why these options? Well, PaaS is quite desirable for developers. And I haven't dug in running Kubernetes (k8s) nor I will. I've got friends and contacts for that. With buildpacks Python -> OCI. - GG - Well, I'm ready to take the challenge, if it needs to be encountered. It won't be easy. It'll be painful, steep learning curve, but that's when you'll learn stuff. Suffer, and get it done! After it's done, re-verify, check and reconsider everything at least a few times.
Linux zswap (@ Wikipedia) decompresses pages before writing them from it's compressed RAM pages to the disk swapped blocks. Why? - Microsoft Windows doesn't. - This just adds amount of I/O bytes. Why decompress on swap, instead of decompressing on unswap or when reading the memory pages back from the swap? - Yes, I've asked from AI, and even before that, I did know "why" it is like it is right now. But the main question is, why it hasn't been improved?
Ah, so much fun with Windows Server AD / Domain & NTFS ACLs AppLocker (@ Wikipedia) and "all the usual stuff". - But not more details about that right now. - Nothing special, just the ordinary stuff.
About EU's (very) slow wake-up to US cloud / hosting related security risks. There's nothing new about it. But it's to wake up late than never. kw: Google, Amazon, Microsoft, Office, Windows, Gmail, and so on, PaaS, SaaS, IaaS
Studied Cloudflare (@ Wikipedia) Durable Objects and Cloudflare R2, Google Cloud Spanner, Google Cloud Firestore, AWS Aurora, Amazon DynamoDB Global Tables and Azure SQL Database, and Azure Cosmos DB for globally distributed applications. - Lot's of options and very different trade-offs to consider, like write latency/availability, is multi-active write pattern support worth of the costs and potential drawbacks?
2025-11-30