310 karma · joined August 1, 2019
The thing is, customers don't necessarily want a distributed messaging system, they might be perfectly fine with a decentralized one if it supports E2EE. Matrix went the way they did and it's far from obvious whether it was a good choice. I'm leaning towards no.
> but someone's done it three times
You can do it however times you want and it will be easy to do if you have access to one of the few domain experts of the protocol. It can be considered easy to implement once someone else successfully implements it and continues to support it for a significant period of time. Rewriting one thing three times inside a single org is not something of interest.
> Blockchains are Merkle trees
This is getting offtopic, but anyway. Blockchains are DAGs, they are not necessarily trees. The canonical chain and candidate blocks have only one genesis block, and non-genesis blocks could have multiple parents (e.g. one main parent and orphaned uncles), which by traverse still converge to the same genesis block.
Although Merkle trees are used in the underlying block data structure, this is orthogonal to the blockchain concept and any data structure could be stored there.
> MLS
Unless I'm missing something about MLS, you're mixing unrelated stuff here.
Technically correct, but in practice the spec is complex enough that there are only 2 more or less working server implementations, both of them developed by the same org. I tried to implement it myself, but in comparison to IRC and XMPP it's just too complex. That's what "effectively" means.
> There are several independent client [3] and server [4] implementations.
Only one non-Element server is there is "Production ready", and it's dead. The website is down and the last commit is from 2021. There were many spec changes since then, including the infamous "sliding window sync", which is necessary for the next gen clients.
> It does not use a blockchain (there is no proof-of-anything). You might be confusing blockchains with Merkle Trees, but in that case you'd also have to think git is a blockchain.
They probably mean the DAG-like structure of rooms, which is a horror to implement, and even more horrifying to implement efficiently. It's a blockchain of sorts, with room events being linked to parent events.
> there is no proof-of-anything
It does not need to be "proof-of-anything" to be a blockchain, that's not why there are words "block" and "chain" are there. There is a consensus protocol between home servers, and it fails terribly, leading to room resets.
This has been my experience with how Matrix handles critique in general, e.g. the response to "why not matrix?" which left me disappointed. Looking at this, and also at low quality of the reference clients and servers, over-engineered protocols and the poor UX in general it does not convince me that Matrix is a good solution to consider.
I don't think it globally matters and it certainly does not scale. Are you saying that Debian devs review all code in their repos? I doubt that, and they're certainly let log4j into their repos.
We already have pretty good automated tooling for detecting known vulnerabilities in deps, and updating and rebuilding is not hard. I don't see the added value of Debian's managing build-time deps of my apps.
It should be the devs' responsibility to audit code they use in their projects (including the toolchain), as well as it should be their responsibility to package, maintain and support the final product. I wouldn't want to support copies of my projects which were somehow modified by the 3rd-party maintainers before being provided to end users.
> the vast majority of projects don't have users on all the mainstream distros, macOS and Windows
I don't need to go far, if we're talking for example about some internal corporate tools, then the vast majority of projects have the majority of their users on macOS and Windows. In rare cases some deb-based usage is supported on best effort, but any other distro -- good luck!
The situation is not much different with some public projects, almost all devs don't go further than "supply a .deb and be done", and it's lucky if they support even that.
> It is great if cargo allows you to say "builds for 50 different OSes", but if nobody ever tested them, to me it's just marketing.
Cargo is not an end-user package management tool, it's a build tool for devs, which is then used for building deb, yum, whatever packages you need. It tracks mostly just build-time deps, and for everything else such as glibc, sqlite, any other .so I have no choice other than apt on Debian. This is fine with me.
> Ideally the distro ships with those dependencies
This is all good and well in the world where there is one single Linux distro, but usually you want to target all mainstream distros and macOS and Windows and what now. Depending on system packages becomes a brittle solution in these cases, since who knows which version of the necessary lib is packaged on this ancient Debian installation. If you depend on a newer version, then you are just forcing you users-developers to either spend time packaging it properly or just running `make install` and littering the system with untracked files. Honestly, stuff like `cargo` fixes this elegantly so I never have to think about it again.
> "just use a monolith for everything" falls apart pretty quickly at larger team sizes.
Absolutely! Don't put everything in a single monolith, develop several mediumliths with lower requirements for coupling code together, but keeping it small enough so that upgrades are bearable.
You are completely right, every project is different, and that's why the general idea of microservices is bad. Split your services where there are technical or heavy political and organizational reasons to do so, but don't go for splitting as your default solution.
> there is a simple way to stop this attack, which is to disable bluetooth
This doesn't work, I've already tried it with my iPhone and a friend's Flipper.
In that case I'll just rollback the first one. I see no reason to logically decouple two processes into two services when the separation is technically driven. They share the queues, the database, the domain, further division will just make things more complex with no obvious advantage.
They do mean that a repo = a codebase. A remark about "set of repos" is just a technicality of distributed VCSs, meaning that each copy cloned from a single source (or from other copies) is a single repo.
Nevertheless, glibc is not the only system dependency which is often used by apps, another such example is JVM of a specific version, which is not tracked by any of the popular JVM build tools and package managers; and OpenSSL, which is too installed by the system package manager.
Dockerfiles exists to fix this by defining all the necessary installation steps as a sequence of instructions, but this is far from a proper package manager.
1. One app - one repo. There is nothing fundamentally wrong with multiple apps (as in separate pods/containers/processes) being developed in a single repo, as long as they are tightly coupled functionally, share the release cycle, but still need to be deployed separately to get advantages of separate processes and independent scaling. The first example that comes to mind is separation of public api and worker processes, e.g. ruby's sidekiq, python's celery or a generic kafka consumer.
2. > A twelve-factor app never relies on implicit existence of system-wide packages.
This is incredibly hard to achieve in practice if you're not running something nix, and even then there are leaks of dependencies on kernel syscall APIs. Every Rust app implicitly depends on a glibc (except musl ones), which is present in major Linux distros except slim ones such as Alpine. In fact, the whole Docker crutch IMO is needed specifically to alleviate this problem.
3. > The twelve-factor app stores config in environment variables
It seems to me much more brittle, as it forces you to store secrets in envs, which could be less secure, and to abandon structured file configs, which may enjoy type safety, IDE autocomplete and automatic parsing. With envs you are forced to implement parses for non-trivial and non-string config vars yourself. In fact, envs are often committed into repos as well in form of dot-env files, so the point about commit-safety is moot as well.
It is a typical situation to observe "Failed to decrypt message" across the whole chat history. It is typical to observe reordering or even lack of messages, which make reading the chat unbearable. It is typical to have problems with session validation.
On one account I have even lost the validated status of another member's sessions, even though we have validated each other before and the only thing that changed is that they logged in with a different client. Since we live in different countries far away from each other, we just have to look at the red shield every time we open the chat and keep in mind that it's not a MITM but just a Matrix failure.
> XMPP just sucks on mobile
So does Element. I have <10 chats joined, and every time I open the Element app I have to wait for 10+ secs before I can start reading new messages. This will surely be fixed by sliding sync, but it's a problem of its own.
On the other hand, XMPP desktop clients certainly work better and faster than Element, although some of them look quite old, which doesn't take away from their functionality.
IMO it's being heavily overstated how Matrix is better than XMPP.
Also a nice read: https://telegra.ph/why-not-matrix-08-07
And the answer https://lobste.rs/s/wvi9xw/why_not_matrix#c_erbsnb