Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)A
Posts
2
Comments
408
Joined
2 yr. ago

  • Deleted

    Virus

    Jump
  • We recently had a funny problem. Our service ran fine, but a postgres upgrade failed because some pg internals were broken (broken ref ids). Dumping the DB also failed for the same error. Reading and writing was still fine, though. So we restored backup after backup... no dice. They all had the same issue: it was working for the service but we couldn't perform any maintenance. Ultimately we had to "manually" dump the data of the service and replay it into a fresh db. That took quite long. But that was interesting, since even the verification of the backups didn't help us notice that kind of corruption.

  • I am also a former TeX addict, but I was always more in favor or ConTeXt over LaTeX. And Typst is basically ConTeXt, but a lot faster (as in you get real time preview as you type).

  • Huh? What's wrong with Overleaf?

    If you "only" need beautiful PDF and it doesn't have to be online, you can also use Typst with vscode and tinymist as editor locally. Not as powerful as TeX, but I know few people for use TeX even remotely to its fullest. The upside of Typst is, that the "core" syntax for content writing is very markdown-like, so you can focus on writing instead of the underlying language.

  • Server written in C++ and client in Java and Lua.... now that's an atypical combination. It still peaks my interest.

  • No, I keep that private to minimize the information I leak about what I host, sorry. (I also don't do git-ops for my server; I back the mentioned directories up via kopia so in case of recovery I just restore the last working state of data+config. I don't have much need to version the configs.)

  • What I did to get rid of my mess, was to containerize service after service using podman. I mount all volumes in a unified location and define all containers as quadlets (systemd services). My backup therefore consists of the base directory where all my container volumes live in subdirectories and the directory with the systemd units for the quadlets.

    That way I was able to slowly unify my setup without risking to break all at once. Plus, I can easily replicate it on any server that has podman.

  • Backblaze B2 using Kopia

  • No, since at the moment it wants to manage certificates, but I don't intend to run pangolin as my main reverse proxy.

  • Pangolin is the most user friendly self hosted alternative to Cloudflare tunnels. There are dozens alternatives, but none with that feature set and such a UI.

  • That would be so damn awsome, if I could finally play 4k 120Hz GfN on Linux.

  • I would rather bet that most people have no clue what an operating system is and that the one they (unknowingly) use is made by Microsoft. On the other hand if they play games (on that PC), they will know Steam, because they actively had to install it and click its icon frequently.

  • Yes. You can simply not expose SMTP at all and just use the IMAP/JMAP part. Unless you need also JMAP, I am not sure it brings you a lot to the table you wouldn't also get from a good old dovecot. IMO the big advantage of Stalwart is the all-in-one package it delivers plus the good defaults. It also shines when you want a multi node deployment. For a single node IMAP only it might not be the best choice, in my opinion. But it would work, if you want to.

  • We can ask, but the indicators are there:

    • it has roadmap with bigger features that slowly shrinks as they get implemented
    • new versions still bring big reworks (I think this is the third time now that the data structure is being migrated)
    • optimizations happen between the versions
    • benchmarks are still on the horizon
  • It aims at both, otherwise it wouldn't ship with sqlite and rocksdb. Stalwarts default is clearly for single node setups and expanding it to clustering takes further steps. So while it supports large scale deployments, it should not be limited to it.

  • It's a 0.x release. It makes sense building the intended features first before optimizing heavily. There's no point having an optimized data structure that then falls flat once you need to add new features that brings new requirements to the data structure.

    Once they label it 1.x (i.e. feature complete and production ready) I would expect it to be optimized. If it isn't, criticism is warranted.

  • The linked ticket also references a merge request that went stale. So I would assume this is a good starting point (I haven't looked at the MR though, so I don't know how far off from the potentially accepted solution it is).

  • I don't think there is a technical reason. Simply no one was interested in implementing it yet. See Nate's answer over at reddit and the associated ticket.

    So once someone is motivated enough it will happen. But without contribution or extreme boredom by the core mainteners (haha) it won't happen.

  • Well exactly as you say: it's a single service instead of having to combine multiple. In my case dovecot was a lot faster for my mailboxes, but postfix was a piece of shit and I was happy to get rid of it and the many components (rspamd, dkimproxy, etc.) it required. It has far too many footguns, and I shot myself multiple times with them over the years. So the most important part (SMTP) is significantly simpler and IMO better with stalwart. And the mailbox part hopefully evolves as well (it already has JMAP, so that is already an advantage over dovecot as well).

  • Use Stalwart as mailserver. Besides coming with sane defaults, it allows to put hooks into almost every mail stage. Those hooks can be sieve scripts, local binaries or http calls.