navi

https://social.vlhl.dev/users/navi

♫♪ never fade away ♪♫

openrc maintainer, works on wine

<3 lizzy مثلية الله ☭ :heart_lesbian_diagonal2: :cuwu: (npub13qr…egvc)
<3 rose (npub146n…l93d)

pronouns: she/her

Public Key

npub1ku39qhads3p8a5tgr5krrwl30239z00v8xzqvnlxjtv6znmewwesr687t7

NIP-05 Address

navi_at_social.vlhl.dev@momostr.pink

Profile Code

nprofile1qqstwgjst7kcgsn7695p6tp3h0ch4gj38hkrnpqxflnf9kdpfauh8vcpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dspy80rc

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1ku39qhads3p8a5tgr5krrwl30239z00v8xzqvnlxjtv6znmewwesr687t7

Show more details

Last Notes

2026-05-19 00:17:22 UTC

- reply

> Why does the player have the information necessary to see through a wall if it should be impossible from their position?
sound effects, which need a source position to be simulated properly, and lag/high-ping mitigations
modern games don't send info when it's impossible for any interaction to happen, but they do send it earlier than strictly necessary so that gameplay experience isn't broken by network latency spikes
the best solution for this though, imo, are small to medium community servers like we had before -- in those, cheaters are much more easily caught by people and kicked out

2026-05-19 00:11:40 UTC

- reply

wayland has a WIP gamepad protocol that would not only let the compositor tell the program what kind of gamepad layout to display on the ui, but also allow the program to be none the wiser wrt to remaps (as they don't see hardware button ids, just button press events)
so in general newer apis are doing it right

2026-05-19 00:06:05 UTC

- reply

only valve and nintendo are trying new things with controllers, tbh -- and valve is hitting it spot on with gyro and dual touchpads
i'm not a fan of a law requiring a specific controller layout at all, everything would become an xbox controller
if we have to have it, then i'd rather it be at most require adaptive icons in the ui like how steam does it, changing the displayed layout based on the controller being used, and nothing else

2026-05-15 01:26:46 UTC

- reply

password store android is a client made specifically for pass, https://f-droid.org/packages/app.passwordstore.agrahn/ -- git, ssh, gpg (via OpenKeychain) and all
pass' major issue is this generic ass name ngl

2026-05-15 01:23:10 UTC

- reply

password store on android works well ime and has ssh/git support
so any platform with git works (aka all of them), the main issue is probably just an economic ui, which idk if one exists for windows or mac

2026-05-08 11:13:39 UTC

- reply

overwriting one entry in the array as a whole seems to be the goal here though?
i think you got the question confused a bit

2026-05-08 01:11:07 UTC

- reply

posix goes beyond just linux though?
that's the whole point, software targeting posix means you don't have to rewrite every bit of software just to use it in a different os
if you want a world where you have more than one option of OS to run your programs on, then improving posix is the way to go, as the alternative is rewriting everything for every new exotic system interface

2026-05-08 00:58:05 UTC

- reply

as long as it's still posix compliant, i'm all for redoing stuff
problem with most osdev things is that they come up with 100% better than before trust me interfaces, and slap a posix compat layer, and things just break or get awkward

2026-05-08 00:36:33 UTC

- reply

iirc those vulns could be used to corrupt any page, so i think they could've been used against preexisting privileged processes too? not sure
but yeah setuid sucks

2026-05-07 01:48:52 UTC

- reply

gentoo is currently also experimenting with gentoo/hurd, it's going well

2026-05-06 01:40:33 UTC

- reply

service namespacing (or sandboxing i guess) is planned for openrc
i've learnt a lot about it lately so eventually i'll take a stab at that

2026-05-03 21:16:03 UTC

- reply

i honestly did not ever need more than subshells in background jobs + `wait`
anywhere where i needed more parallelism than that (which, really rare), i'd say shell wasn't the right language for the thing as a whole anyway

2026-05-03 13:11:30 UTC

the point where that's needed is the point i'd say the tool shouldn't be written in shell, or should be simplified over multiple tools and then those go in a shell script
#note1axt…3yf2

2026-04-28 14:20:44 UTC

additionally, my phone has internet access as normal, it's just my laptop??
#note1j8m…na3n

2026-04-28 12:18:30 UTC

how do i debug "whenever wifey joins the minecraft server on her laptop, my laptop's internet (both wifi) slows to a crawl"
it goes back to normal as soon as she disconnects, and i have no idea where to even start with figuring this out

2026-04-24 15:36:12 UTC

> project about distributed git-based bug tracking
> uses github as the main contribution and distribution platform
huh

2026-04-23 17:20:26 UTC

> Minutes of the 23rd April 2026 Teleconference
> Bug 1976: specify array variables and syntax
https://www.austingroupbugs.net/view.php?id=1976
> The committee agrees that this is a useful addition (for Issue 9), but we need a lot more words than currently proposed in
> Desired Action.
> ACTION: Andrew to contact submitter to ask for some more words.
> ACTION: Eric B to request The Open Group to sponsor this change once we have words.
HOLY HOLY HOLY HOLY
LET'S GO, POSIX SHELL ARRAYS

2026-04-22 21:04:55 UTC

- reply

depending on what "countertop" means, even the savings from water along might just save up money -- as most dishwashers use very little water to wash a bunch of dishes, compared to what most people would use washing by hand

2026-04-21 23:06:55 UTC

- reply

gentoo doesn't have a "installer", the live cd is a linux environment for people to do the install on, but it doesn't contain any installation files -- you still fetch the stage3 from the web
so if you want to install gentoo with just cds, on one cd, put whatever linux distro you want, on another cd (or the first one if there's still space, i guess), but the stage3 tarball, it's about 250 megs
boot the first one, swap out, extract the tarball, continue the install
and good luck actually using that machine for anything without first copying a dozen packages via cds as well

2026-04-21 23:01:51 UTC

- reply

from another machine, which is kinda obvious with the context of "air-gapped systems"

2026-04-21 18:20:57 UTC

- reply

myea, (i think? not sure now) i knew that
though where i currently live i don't have a lot of choices for coffee beans anyway (nor the spoon to figure out which one is good)
ftr by black coffee i meant without milk or other volume additives, not really the roast (but whenever i move out and have the chance and energy to try, i'll try buying decent light roasted beans)

2026-04-21 18:16:57 UTC

- reply

as some sort of personal report, coffee for me would work kindof sorta work and only if i drank like 3 or 4 cups of milk-coffee a day (black coffee would have more caffeine but it's too bitter)
(tho now i have lisdex, which kind of sorta works by default, and actually works if i take two 30mg pills instead of one, although that's not what's in my prescription)

2026-04-21 15:09:31 UTC

- reply

you can install gentoo from any linux distro, you don't have to use gentoo's minimal cd -- 90% of the installation is done chrooted in after fetching stage3

2026-04-21 14:31:35 UTC

- reply

the gentoo install cd was under 700MiB until about a year ago i think -- now it's 1.05GiB as iirc they decided that having more QoL tools built-in was more important than keeping it cd burnable

2026-04-21 14:18:42 UTC

- reply

for context, in bash, /bin/sh, or zsh with `-o shwordsplit`, if you did `search-files "name with spaces"`
it would become `grep -rbw . -e "name" "with" "spaces"`, or, searching `name` in the directories `with` and `spaces`
the quote makes splitting on spaces not happen, so grep gets `"name with spaces"` properly

2026-04-21 14:16:27 UTC

- reply

please quote "$@", zsh """protects""" you by disabling word splitting by default, but even then building the habit of properly quoting variables when you don't want splitting is important imo

2026-04-20 18:56:02 UTC

the only thing i actively dislike about vulkan's api is sType tbh
i understand needing to set sType on .pNext structures
but why the hell do i need to set `.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,` in a `VkCommandBufferBeginInfo` that i'm passing to `vkBeginCommandBuffer(cmd_buf, &begin_info)`
like seriously vkBeginCommandBuffer takes a typed pointer, there's no reason for me to need to type the whole tag
and this wouldn't be such a big problem if not for:
` $ grep '\.sType' vulkan.c | wc -l > 42`
gigantic long names, that you have to type over and over, for no practical reason
the main thing vulkan messes up is not providing helpers for those structs, really, helpers for common defaults would go such a long way to make it more approachable for beginners (there's only a third party util library for helpers, in c++, that is c++ only)

2026-04-20 01:01:15 UTC

- reply

honestly in my experience anything fuse that does the same as something in the kernel, ends up kinda abandoned because everyone just keeps using the kernel impl
which is a shame

2026-04-18 17:43:53 UTC

- reply

> That makes it Linux (or POSIX) specific though which kind of sucks. The agnosticism was a feature to me.
not using posix apis is a downside to me, because of the mentioned "any program that has to interact with it, now has to implement two filesystem apis", and if you want it to work with shell programs then you have to mesh both filesystem views together or make special syntax and it's a mess
note that hurd has RPC libraries that can be used, and the libc posix apis are built on top of those -- which i think is a good design as you don't have to handle separate filesystem namespaces on (most) tools, which were never designed to do so, but anything that is okay being not-posix, can use the rpc libs
> It is but it gets rid of the agnosticism (in this case).
it doesn't, not by design, you just need somewhere to route things, that's why microkernels usually have a standalone "virtual file system" daemon (or sometimes it's in kernel, with filesystems proper as daemons)
programs make the rpc to the vfs, that then dispatches it to daemons -- linux fuse kinda works the same way, but using the in kernel vfs
what to me you seem to want, is that but over a non-posix apis, i don't quite relate but the choice of ipc isn't tied to the backend model, if your design is half-decent
> that’s not just dumb but also intentional:
i can try opening an issue later with them, because it's really stupid so maybe if i just explain it, they'd change it

2026-04-18 17:23:21 UTC

- reply

> I'm not sure we're talking about quite the same use.
i really though u were talking about:
> yes, fusermount3 which does the eventual callback registration does
so, sorry!
---
but davfs2 for some inane reason exclusively distributes the /usr/sbin/mount.davfs helper and nothing else.
that's not just dumb but also intentional: https://github.com/alisarctl/davfs2/tree/main#2-security-considerations
davfs seems to not understand how fuse works at all, because that whole section is stupid and would only apply to normal non-fuse filesystems, like:
> To prevent inexperienced (or even malicious) users from introducing dangerous content into system directories or other users home directory, the administrator must have control over user mounts.
do they not know that rootless fuse just won't let non-root users do that...?? you need permissions over the mount point
though the unshare thing would work for this as well
---
> But yeah, the idea of not having to implement a full special snowflake FUSE daemon for every single protocol and instead providing protocol implementations to some IPC/RPC manager (that can accept backend-specific configuration options as necessary)
if you see the kernel as ipc via syscalls, the fuse daemons are your backends -- the architecture doesn't change much at all but thanks to dbus you add a *lot* of overhead
having a daemon per filesystem supported is imo ideal, or even better one instance of it per backing device (be it a physical device or a networked resource) -- iirc that's how hurd does it, not super sure, but while porting openrc to it, i had to set translators on filesystem nodes and those are basically started automatically as normal system processes
my own osdev/kernel project will be similar, daemons for filesystems, and a daemon for the vfs as a whole, libc sets up ipc calls between them, kernel proper does nothing except scheduling tasks (not even spawning processes, i wanna do that in a daemon too)

2026-04-18 16:45:26 UTC

- reply

> This started as issues with davfs2 asking for root permissions despite being a FUSE filesystem (defeating much of the point of using FUSE to start with).
it only needs root to call `mount`, the fs process itself runs as your user
this is because the default mount namespace requires it, but it's not *hard* required, e.g.:
```
navi@orchid ~
$ doas chmod u-s /usr/bin/fusermount3
Place your right index finger on the fingerprint reader
navi@orchid ~
$ aft-mtp-mount mnt
fusermount3: mount failed: Operation not permitted
navi@orchid ~
$ ls mnt
navi@orchid ~
$ unshare -rm zsh
root@orchid ~
$ aft-mtp-mount mnt
root@orchid ~
$ ls mnt
'Internal shared storage'
```
(but this ends up still requiring setuid for the newuidmap(1), for mapping the root user -- i've been told there's a way to setup a unpriviledged mount namespace without suid helpers but i didn't yet figure out how)
> As for FUSE, I just saw no reason to involve the kernel at any point (including registering callbacks for a userspace program).
well if you want `open`/`read`/`readdir`/etc to work, you kinda have to do that, otherwise we'd need every tool that interacts with the fs to implement alternative filesystem apis specific to whatever implementation of the fs there is
and imo that is a no go for like, implementing "usual" filesystems in userspace instead of in kernel
really the problem here is monolithic kernels being a not-great design :blep:

2026-04-18 08:18:54 UTC

- reply

filesystems being in kernel is kinda dumb yea
but without any kernel interaction sounds like a pain to synchronize everything
what's the issue with fuse here again, i forgot

2026-04-18 08:17:51 UTC

- reply

ipc is good, yes
the problem is that dbus isn't

2026-04-18 08:08:20 UTC

- reply

ever since i learnt i couldn't cross compile a desktop solely because glib, i've started doubting anything in that ecosystem (not to mention how glib is OOP C full of UB)
ftr cross doesn't work in glib because they want to build a program for the target machine, and run it during build to introspect apis, and have no support for doing it in cross like proper codegen tools would

2026-04-17 20:42:18 UTC

- reply

people don't seem to remember when being a hacker was an act of activism and rebellion
rebel, protest, fuck this abusive shit

2026-04-13 19:46:38 UTC

- reply

unless you have formally proven all your daemons, and is 100% sure that the hardware is functioning perfect and all your input is what you expect
then you can't know for sure it won't ever crash, yes a system that doesn't crash is reliable, but a system that reliably doesn't crash doesn't exist -- thus we have supervision, so we can recover from crashes, and achieve reliability

2026-04-13 18:21:44 UTC

- reply

system-v unix (and earlier, i think?) having /etc/rc be a shell script that is just executed, bsd followed suit, kinda -- you just add execution of the daemon binary in there
if you're not considering that daemons will crash, and not thinking much about stopping them, then just having them start and then fork themselves off makes sense
so in an utopia of computing where nothing crashes unexpectedly, it might be simpler to just let daemons be by itself, not in our world though, but it took a while for people to learn that (and some people still didn't, claiming that instead of supervision we should "just" fix all crashes, who needs reliability anyway??)

2026-04-13 15:38:53 UTC

- reply

older systems (and sadly some modern ones, as much as i'm trying to stop that habit in openrc, and most bsds still do this too), would have daemons start, fork to the background, and write their pid to a file, usually in /run
and then when you tell the init script to stop the given service, it would read the pidfile, (hopefully do some sanity checks), then send a TERM or KILL signal to that process
but then what happens if said daemon crashes, the pid wraps around, and something else is started under the pid written to that file? with luck the sanity checks make the operation fail, without luck you kill a random unrelated process
modern solution is called supervision, you run the daemons directly under a supervisor, without background forking -- so that you don't have to rely on a file that might get stale

2026-04-13 15:33:42 UTC

- reply

yes, though not likely an overflow, but once it reaches kernel.pid_max (see https://social.vlhl.dev/notice/B5FqXqKZfAwq8VVpc8), it wraps around to the lowest unused pid
which is why pidfiles are *bad* for daemon management

2026-04-13 14:43:27 UTC

pid: 2148769
i think my computer has been on for too long

2026-04-13 14:38:08 UTC

- reply

the goal was to not use the mastoapi tho =w= (impossible task)

2026-04-13 14:32:30 UTC

- reply

i read the ap spec and went "there's potential but fuck how do i even log in?" (context is i want c2s)

2026-04-13 13:06:13 UTC

i got reminded teen titans go exists (and what the fuck it has 400 episodes now)
literal trash, but now i wanna re-watch the og teen titans tbh

2026-04-11 22:52:09 UTC

- reply

that still doesn't make up for the pain that it is to interact with for literally anything except writing code

2026-04-11 22:30:45 UTC

why the hell does `uv` takes 5 minutes to build on a modern amd cpu with -j16
it's a build system... and took all that time while pinning all my cores to 100%

2026-04-05 21:34:50 UTC

sometimes i wish people would use the c type system properly
instead of "everything is number"

2026-04-05 20:01:35 UTC

i still don't understand why systemd is pushing for /efi, when like either /boot, or /boot/efi, make so much more sense

2026-04-05 18:52:44 UTC

- reply

a) have some process with higher privileges open it and send a file descriptor over (seatd approach)
b) add the user to an acl (elogind) or unix groups (traditional)
i personally think unix groups are fine and limiting user access to devices is the wrong way to go about it, as the proper thing to do would be to control which programs/processes are allowed to do $thing, not which users

2026-04-03 20:17:25 UTC

- reply

i am completely lost because the transport protocol has nothing to with priviledge

2026-04-03 20:07:21 UTC

what is a good soldering iron
is the pinecil any good, or what should i get

2026-04-03 20:03:28 UTC

- reply

if the thing can be executed as a script, then nothing stops it from also handling cgi
there's no issue here inherently to cgi, except that people that like cgi are usually the people that don't like over reaching frameworks

2026-04-03 19:49:17 UTC

- reply

what?
cgi works by giving the process environment variables, inputting data on stdin, and reading the response from stdout
there's a few cgi libraries that basically just do http/html stuff and print it, so, i don't understand your statement

2026-04-03 19:45:22 UTC

- reply

"javascript has frameworks that work via http" is valid to say
"$language has $framework that works via the cgi" same thing

2026-04-03 17:11:41 UTC

- reply

i mean the coordinates on windows isn't the corners, but the sides
so, i'd cry to have brought a parallelogram monitor

2026-04-03 17:01:09 UTC

windows using 4 coordinates (top, right bottom, left) instead of 2 coordinates and 2 extents (x, y, width, height) is so painful

2026-04-03 14:57:22 UTC

- reply

i had no pfp at all for like a week

2026-04-03 03:18:24 UTC

i'm tired of being blobcat_comfy
temporary pfp, but have a pfp again

2026-04-02 02:59:41 UTC

- reply

as of posix 2024 -e and -o pipefail are both part of posix sh
-u is the only bashism now

2026-04-01 23:05:31 UTC

openrc's `reexport` vars are now also being reexported by pam for user services!
so on the next openrc version, doing `rc-update --user add dbus boot` will be all one needs to do to have dbus working *and* the variable properly set on your login environment!
only applies for services in the `boot` runlevel, as some sort of special semantic for that runlevel -- having it mean "synchronization point for core services of the {system,user}"
not sure if i am a fan of the semantic or not, but it's the cleanest i came up with

2026-04-01 14:26:56 UTC

https://www.gentoo.org/news/2026/04/01/gentoo-hurd.html
lesgo :3

2026-03-31 01:33:45 UTC

- reply

i never vibed with debian packaging
gentoo it's like, drop a file in a repo that consists of half a dozen bash variable assignments and maybe a shell function, now `emerge` can install it
debian is this weird multifile thing that can only be done easily with n helper tools and you have to build the package by hand to test it's just aaaaaa

2026-03-27 23:29:01 UTC

i think the fact that there's people that 'look up' to me will always be a bit of a wild concept
like what do u meeeeaan i'm just a gal that yaps too much, has waayy too many ideas and opinions, and about double as many unfinished projects
=w=

2026-03-24 00:13:36 UTC

oh, posix 2024's `read` had -d for a delimiter
that makes a lot of things so much easier to do, wtf
love it

2026-03-22 13:23:46 UTC

saying "sideloading programs" in a linux desktop context is wild, please fucking don't
there's not supposed to be one and only "official" loading, where everything else is "side" to it
it's called installing programs

2026-03-20 20:21:55 UTC

- reply

i mean cgit is a cgi tool so it's executed anew for every request, so it couldn't really keep it in memory
though i guess one could put the cache in tmpfs and that would make it fully in-memory

2026-03-19 10:12:12 UTC

https://social.vlhl.dev/media/84a9d8673d0a409817c0aaa2e4c490588759a2c884e166c689965f8aca2395bb.png

2026-03-18 09:02:27 UTC

"but, what does 'LLM code' means??? and how would we police it???"
it... ain't that hard fam, we're not trying to solve the philosophical conundrum of sentience
occam's razor applies, the simplest explanation, it's "code that was generated by a large language model" and if that's not clear enough then nothing is, and words are meaningless
how do you police it? well how do you police people sending plagiarized code? do the same for LLM code, if the project policy says no LLM code, assume people to respected that unless proven otherwise
i'm so tired of this argument that basically just tries to derail from the actual topic

2026-03-14 17:46:45 UTC

i wonder what i could do, if anything at all, to get better openrc support on base debian and base arch

2026-03-14 11:50:04 UTC

- reply

"it's just a tool" and "we're neutral" is the new "keep politics out of FOSS"
this is hell

2026-03-14 10:42:24 UTC

https://github.com/ImageMagick/ImageMagick/discussions/8615#discussioncomment-16129293
here we go again
#note1ujp…0zd7

2026-03-13 19:36:40 UTC

https://github.com/ImageMagick/ImageMagick/issues/8614
well, i'm trying

2026-03-07 10:29:53 UTC

- reply

posix shell :3

2026-03-03 00:00:35 UTC

damn guilty by... unknown association?
and who even are those instances, not once have anyone from their team reached out to ask if we're even aware of that
#note1lar…fajl

2026-03-02 20:33:59 UTC

i am not surprised, slightly disappointed but this was coming anyway
#note1jrn…73uv

2026-03-02 15:13:09 UTC

and here's why "language specific" unvetted and user-driven repositories are hell, and by extension why distro(-like) packager-driven package managers are good
#note17lp…2fpl

2026-03-01 12:02:45 UTC

i hope the original post is a joke, but also it's kinda funny how some people actually believe that this is the only way c should be "accepted" or smth
#note1py2…5xzp

2026-02-27 14:35:47 UTC

- reply

i'll stick to it because it's pretty much the only solution for multi-user synchronized calendaring, and supported everywhere
storing each VEVENT/VTODO/VJOURNAL as it's own file simplifies a lot, and ical being flat means i can stream it in and deconstruct it as i go

2026-02-27 14:30:36 UTC

- reply

i should actually read the wire protocol of caldav huh
i knew radicale keeps them in disk as separate files, but the endpoints were giving me a single ics per calendar
maybe it's the common pattern to store them split, merge them for transport, then re-split

2026-02-27 14:16:11 UTC

the most ???? design choice in ical for me, is, why is the whole calendar a single file ~w~
could've made the calendar a directory, and events be files, blep

2026-02-17 23:15:59 UTC

how bad is it to create a new fedi account with the same username in the same instance as a previously created-then-deleted account?
asking because fedi is fucked up and prone to blowing up

2026-02-10 18:08:54 UTC

how do i make a gtk theme that just... changes adwaita colors
if i define the colors and nothing else on $theme/gtk-4.0/gtk.css, everything is now transparent
i assume i have to import adwaita somehow but i can't find out how

2026-02-07 17:12:48 UTC

- reply

i could look into it, though i have doubts if it would a) work well for gaming, and b) work on a non-dark room (last time i used a projector was about a decade ago...)

2026-02-07 17:09:25 UTC

- reply

i live with wifey, the main point of the tv is to watch things together -- if i lived alone i'd probably be fine by sitting in a desk with a usual computer monitor

2026-02-07 17:05:54 UTC

i still hate the non-existence of modern non-smart TVs
i don't want a built-in network capable computer on my tv running some shitty proprietary android interface
computer monitors aren't really an option because a 55" computer monitor either doesn't exist or costs over a thousand bucks

2026-02-05 13:26:41 UTC

i wanna have a home server, and a few diskless thin clients on things like tv, kitchen, room -- the client would ipxe boot from rootfs' built by and kept by the server
the tv one would also play games out of the server via waypipe and such, all wired via ethernet in the walls
does anyone have recommendations for the thin clients? i think some arm systems would be ideal but i never built such a board before so not sure where to look

2026-02-03 13:05:47 UTC

https://github.com/OpenRC/openrc/pull/942
this merge request, if i look at it while signed in, i see two contributors, me (author), and @npub1c96…0k8u who gave feedback
but if i open the page logged out, copilot pops up there to the list of participants
for the record openrc does not make use and won't ever make use of copilot or other llm codegen tools, and not once did we get any kind of copilot reply on this merge request
THE FUCK MICROSOFT?
https://social.vlhl.dev/media/6f1da9a085f914f1543b6f14f7c72bc972f414be481ef4e6155994c37fd653c1.png
https://social.vlhl.dev/media/8ebd322962430da5905636afa0d7795d8acda8f2f6ff99ce9fdda8e8189a2705.png

2025-10-07 01:38:40 UTC

- reply

between imap and the "new replacemente" jmap, i take imap over jmap anyday

2025-10-07 01:34:52 UTC

- reply

push email is a thing and was made mainly for mobile platforms to avoid needing the constant polling for new mail: https://en.wikipedia.org/wiki/Push_email -- and is used by thunderbird android (if the option is enabled), it's also used by some desktop clients like aerc to avoid polling since that is shit for performance too
it's not as good for batter as the centralized notification thing google has but that's proprietary and requires custom backends and stuff so

2025-10-07 01:33:21 UTC

- reply

smtp is just only for sending email, to access your mail box, it's done via imap
"push" as thunderbird is probably refering to IMAP push which does indeed keep a connection open if you toggle that option: https://support.mozilla.org/en-US/kb/configure-push-email-thunderbird-android "When using Push, Thunderbird maintains a connection to the mail server"
it does keep a connection, but it's not gonna be actively polling it

2025-10-03 21:33:06 UTC

funny how my instance on global variables and such flipped the more i work with unix code
small, self-contained utilities gain a lot in simplicity by keeping global state as global variables

2025-08-30 00:54:47 UTC

- reply

rust dependencies cannot be dynamically linked, cargo simply doesn't support that (i've tried for gentoo, there's no way, and no interest because that would require a stable abi and rust folks are afraid of committing to one), so distros don't dynamically link it because they can't
ghc on the other hand supports it first hand, so i'd expect it to be more stable :/

2025-08-30 00:27:56 UTC

- reply

updating a package, any package, twice a day is just stupid (bar exceptions like an important hot fix), and if that happens often to multiple packages, it leads me to believe it's automated and barely tested, which puts in question arch's packaging policies, but those reasons have nothing to do with haskell itself (even though they're brought forward by how the ecosystem is)
if you have 229 dependencies, you have 229 dependencies, and hiding that and/or mitigating the frequent updates onslaught by special-casing those dependencies (and vendoring them no less) is treating the symptoms instead of the cause -- and imo such mitigations are kinda bad and only really hide away problems that need fixing

2025-08-30 00:19:40 UTC

- reply

i mean a dependency is a dependency, if a project has a ton of them, why is it bad to have each be a package by itself?
(or am i missing something else that is wrong with their packaging)

2025-08-12 01:34:08 UTC

- reply

that's my experience with like 90% of fedi drama/meta/idk
i only ever get the afterblows in my timeline, or maybe, rarely, a callout post, and then need to either ask people "what the f**k happened", or just decide to not care
more often than not i decide to not care and continue posting about tech tbh

2025-07-26 16:20:11 UTC

- reply

my problem with "making a vcs" is that, well, the whole dev world is git, trying to make something new (or use anything else really) basically ends up reducing contributions by adding a barrier of "downloading and learning this new tool"
any new vcs needs to be compatible with .git and not do anything extra that can't be emulated with git... that or fight years and years of traction ig

2025-07-26 16:07:49 UTC

- reply

"i hope you like parsing text from stdout" kinda pain
data is in structured format in git's process, serialize to stdout, de-serialize on your application
why, exactly, who knows

2025-07-26 16:01:05 UTC

- reply

interacting with git by spawning a git-<applet> instance from anywhere but the shell just sucks

2025-07-26 16:00:34 UTC

- reply

no just the ones that have a common / standard-ish protocol and other programs might want to interact with
i.e. openrc has librc for interacting with the service manager, my fedi impl is based on my own libactivity, and so on

2025-07-26 15:55:11 UTC

another option: implement the whole git cli based on libgit2, and help improve libgit2 as a whole
it's kinda silly that git wasn't a library to begin with
#note1tuj…q7du

2025-07-26 15:36:58 UTC

would it be stupid/petty to fork git if/when they start hard depending on rust
my brain is a mess