L
> […] but if the issue drags on (as the systemd did, and as I suspect the AI one will as well sadly), then it's not clear what holding the vote solved or what it meant. If nothing else, it tells you that there's no solid consensus yet and people's opinions on the subject are unclear. While unfortunate for deciding a direction, it's still a truth everyone can reflect on when interacting with one another. (Better than each camp trying to claim that opposing people are just being loud and really…
L
> The second camp want contracts that are guaranteed to run in release, so they care that they run only once (for perf reasons, and possibly to control eventual side effects) so they are not happy with that approach. While it doesn't guarantee that they run in release (and why is the *developer* telling the *user* what they are and are not allowed to do!), surely if they're worried about performance doesn't that approach mean they *must* run on *every* invocation? Whereas if it runs in the…
L
This is actually a major downside of Rust. A "tree of values" is not a universal basis for a "good internal architecture", nor is a DAG. Often (certainly in network protocol implementations) you *must* be able to index objects in multiple ways. E.g., in a protocol that allows different nodes to publish and re-publish data-objects (a common pattern in networking); - given a neighbour, you must be able to retrieve the subset of data-objects which were received from it (e.g., so you can clean up…
L
But it wasn't just sysv-init vs systemd, it was systemd vs the choice of upstart (whose development was effectively stopped due to the second vote) or openrc. I wasn't advocating against voting, more it is the issue of what is being voted for (e.g. what actually makes sense on the ballot). Looking at the past ~10 years of votes in Debian there are some trends: most "administrative" changes/votes have at most 2 or 3 options, and all are above "further discussion" (and it seems more of a wording…
L
In the extreme, a terminal supporting tmux Control Mode can do all of this, because tmux does the multiplexing and sharing of state, and the terminal does everything else.
L
It should be possible - note the existence of the tmux Control Mode which is intended for use in the other direction (where a terminal emulator controls tmux so that tmux does the multiplexing, and the terminal emulator matches tmux's layout of panes and windows).
L
I got annoyed that Rust wouldn't let me do some things that I could have done in C. BTW, Rust is really good at forcing a good internal architecture, having a tree of values instead of a sea of objects. I sometimes say A DAG of components instead of a spaghetti architecture.
L
I see. Humans shape tools and the tools shape us in return :-)
L
Terminals can be modified. In the simplest case, you can open a new terminal tab and just dump the scrollback.
L
Basically a different programming paradigm that I'm not (yet) used with. It's a question of familiarize myself with a different way of doing things, like contributing to a C project using a very different coding style. At first we find the new thing ugly, and after a while the pill is swallowed. Fair enough. I had a "trying to write C in Rust" period as well while I was learning Rust, and I got annoyed that Rust wouldn't let me do some things that I could have done in C. (One of those is still…
L
What about naming the first hash variable or function argument something else That would still leave it available when you no longer need it, unless you use drop. You'd be avoiding a useful protection mechanism. Old K&R C used to require some things to be globally unique that no longer need to be, such as field names; this is how we ended up with fields like tv_usec rather than just usec. (Technically, field names could be the same if they had the same type and offset; yikes.) That was an…
L
I was talking about the combined case. Yes, if you run zellij-in-zellij, you'd had the same problem, unless zellij starts using some protocol for natively handling scrollback remotely.
L
Are they though? Wouldn't it be one local zellij for the terminal and a remote one for the multiplexer/detach part?
L
How else do you expect that to work for re-attaching and multiplexing though? It is not like terminals have a way to delete and inject scrollback.
L
There probably is a reason why distributions do things such as giving a new version number to libopcodes all the time, or providing a non-PIC libopcodes.a. So vendoring libopcodes is probably not easy. It may be easier to vendor the libc dynamic linker, or possibly even easier to statically link gforth. The situation makes me wonder how this all works out for other programs that use libopcodes. I see with ldd that in Debian 13 objdump dynamically links libopcodes, and gdb does not, so it…
L
If the user can run something that listens on a port, "security policy" clearly already allows that, no?
L
You can compile zellij with this feature completely disabled. There is a "noweb" binary among the project releases too.
L
Rust, C++ or Perl have more operators. In C, creating a range would be done by calling a function. Then it's just a matter of naming it well, with the right parameters, using maybe an enum type to specify the wanted range type (inclusive or exclusive for the last value).
L
Basically a different programming paradigm that I'm not (yet) used with. It's a question of familiarize myself with a different way of doing things, like contributing to a C project using a very different coding style. At first we find the new thing ugly, and after a while the pill is swallowed. Otherwise, I said in a comment above that there is always typedef in C.
L
Refactoring a verbose C codebase (or Java) involves more work, indeed. The IDE can help. In C there is `typedef` to have shorter types to write. Also by splitting a big source file into smaller and more manageable source files/classes. So instead of having long types to write with generics and other stuff in it, you can create a small class to encapsulate the thing. It's a different paradigm.
L
> I'm less fond of the succinct syntax Could you say more about what you mean by this? This is the first time I've ever heard somebody call rust "succinct" as a negative. I'm always looking for feedback on the language from people who have different experiences with it.
L
> a compressed stream of the various channels (e.g., USB, DisplayLink video, audio) over a communication channel (USB 3 in this case) and decode it again for output to the user's devices. Compuserve provided its users with multi-stream transport over a serial line forty years ago. Simultaneous file download/upload, interaction with email user agent, multi-user text chat, even voice. The hardware was dial-up telephone lines, usually with a PC-AT plugin modem card at 14.4 kb/s upto 53 kb/s. The…
L
Wouldn't it be easier to just have the kernel generate and sign the key before userspace runs at all, cap it and keep it around until hibernation time? Just use whatever PCR measures the kernel itself. (9 for for grub, 11 for systemd) For hibernation I can't see a need for decrypting it on any other kernel than the one which actually hibernated.
L
Yes, that sounds good, but io_uring should also be abstracted. For most uses, either an event loop library or a poll-like interface is less error-prone than working with shared queues, and at least for the case in point syscalls are not as much a problem as being forced to launch threads to perform a non-blocking operation is. io_uring is A solution and in the right direction, not the only valid one.
L
I'm not expecting anyone to be bound to the decision, only noting the compromise did not hold, and the issue ended up finally resolved by one group "winning" and the other group leaving. There is then the question of whether the compromise was worth it if the question was never resolved, and whether it would have been better (for everyone, those that agreed and those that disagreed with the final decision) to have avoided the compromise and instead required a (more binding) definitive choice,…
L
Remember the days when we used to build systems out of pluggable components talking to each other over generic interfaces? […] Couldn't it therefore behave like a zoned block device via DM instead of having to be integrated into a specific filesystem? Now I cannot talk about XRAID; I haven’t looked into it, and maybe it’d easily work the way you envision it. But as a general answer: because ZFS, but also others like APFS etc, have realized that block layer abstractions are comparatively crude…
L
The problem is that we do not have a good interface for this. Sure if you just do redundancy and maybe encryption it works well, but even adding integrity creates problems, because we now suddenly need a journal which we already have at the filesystem level, this leads to write amplification. This gets even more problematic when adding deduplication or compression because now your block device does not have a fixed size anymore. The current solution for this is guesstimating the possible size…
L
Hello! After I had my incidents related do bad crawlers I decided to do something about it. I tried with IPs and with user agents and I failed. So, I created AI-drop: https://rocketgit.com/user/catalinux/ai-drop A short presentation is here: https://rocketgit.com/user/catalinux/ai-drop/artefacts/do... I will answer any question / suggestion you may have. Thank you very much for your time!
L
As a proof of concept, my idea with SubC was for the compiler to treat certain library functions specially. Instead of syntactic sugar, having just functions to call, which would follow the Rust culture but with a verbose approach. (A standard C compiler would understand the code as well). In other words, "C-ify" Safe Rust concepts (and culture, like you explain). Emmanuele Bassi from the GNOME community has put together a strawman proposal for this. A small proof of concept would be doable I…