Sadly, the SVG library Nanosvg is now slop.
We embed a sane version in #Dillo to read Wikipedia math which will be kept LLM-free. Human-made changes will continue to be cherry-picked from upstream.
Remote
Dillo browser
@dillo@fosstodon.org
mastodon 4.7.3Dillo is a fast and small graphical web browser with HTTP, HTTPS and FTP support.
Other protocols like #Gemini #Gopher and #Spartan are available via plugins.
⚠ Do not use dillo . org, is not under our control, see: https://dillo-browser.org/dillo.org.html
PGP key: 32E65EC501A1B6FDF8190D293EE6BA977EB2A253
1731 Followers
0 Following
45 Posts
Joined January 14, 2024
Website:
Liberapay:
Website mirror:
Boosted by @trending@homestead.social
I have prepared a directory of sites that work well with #Dillo with a focus on human crafted sites to help new users find useful content:
https://dir.dillo-browser.org/
There is A LOT of content to explore, so be mindful of the time when entering the rabbit hole.
You can also suggest new sites to be added at the end of the page.
Added experimental support to solve go-away and Anubis challenges to the "Fix page" script with a small C helper program to compute the proof-of-work.
You will need to enable session cookies in ~/.dillo/cookiesrc (DEFAULT ACCEPT_SESSION) so that the challenge cookie is saved.
New link navigation mode using the keyboard (coming soon).
Use n and N to select a link from the viewport and press Enter to open it. The hjkl keys are also remapped to navigate around, similar to vim.
#Dillo 3.3.0 finally released! 🥳 😎
Dillo 3.3.0-rc1 released 🥳
Let us know if you encounter any issues to fix them before the 3.3.0 release!
Added support in the "Fix page" script for GitHub issues to read the comments directly from Dillo. They are buried in the HTML itself as JSON.
Please migrate away from GitHub if you can, this is awful. One option is Codeberg, which renders the issue pages in plain HTML.
Introducing the Dillo Lab, a new section in our website for experimental ideas and features.
Just fixed a big blocker for FLTK >= 1.4.
Blurry fonts were caused by a half pixel misalignment bug in FLTK which broke the font hinting leading to poor readability. Should be fixed in next FLTK 1.4.5 release.
I was not expecting to having to follow this whole font rendering pipeline rabbit hole. I'm surprised how little information there is on the Web about this topic, maybe I write more details later.
Replying to
Kind reminder that our #Dillo git repository is distributed among three different sites, all in sync:
- https://git.dillo-browser.org/dillo
- https://codeberg.org/dillo/dillo
- https://git.sr.ht/~dillo/dillo
Currently #Sourcehut is fighting a DDoS and the service is spotty: https://status.sr.ht/issues/2026-04-06-ddos-attack/
This is precisely why we have chosen a replicated set of git servers.
Hi @brewsterkahle@mastodon.archive.org,
Is there a reason why the Archive.org (also searching the Wayback Machine) cannot be used without JS? It excludes people with older computers from using the site as modern web browsers require a lot of resources.
It used to work in the past, but now it doesn't anymore. Could you consider keeping at least the search and download functionality without requiring JS?
Thanks!
#Dillo on the Luckfox Pico Mini board (ARMv7/RISC-V CPU with 64 MiB of RAM), found on YouTube:
#Dillo is also mirrored by the Software Heritage non-profit (@swheritage@mstdn.social). Here is the source code:
And our issue tracker (perks of plain text):
Along with our PGP signed commits, it should have the same level of trust as our own git server.
New commits are now signed with my OpenPGP key so they can also be trusted when cloned from other git mirrors.
https://keys.openpgp.org/search?q=32E65EC501A1B6FDF8190D293EE6BA977EB2A253
Replying to
From the orange site, "Strict grammar for declaring documents is merely a fetish".
Isn't that the best name for a document format to be load in the #Dillo browser?
"If you are a normal visitor and you expected actual content instead of this page, then use a normal, up-to-date browser. It's that simple. [...] You'll have to use another, normal browser. I'm sorry for the inconvenience."
Is 2026 the year of accurate bot protection?
On forking the #Web:
- How do we make sure it doesn't end up like this one?
- Are you willing to pay directly or you prefer to continue paying indirectly via surveillance?
- Why leaving older machines out wouldn't be considered discrimination?
- What are your accessibility favourite DOIs?
Treat for thought: https://sessa.bandcamp.com/album/pequena-vertigem-de-amor
Replying to
@SRAZKVT@tech.lgbt in any case, for this initial approach I'm only considering structure not presentation.
Ideally it should be possible to output three formats: screen, print and non-visual output.
It is probably a good idea to start from the non-visual output and work backwards from there, so accessibility is guaranteed by design.
The nice 0x0.st paste service has disabled new uploads:
% curl -F 'file=@/dev/null' https://0x0.st
uploads disabled because it’s been almost nothing but AI botnet spam for the past few months. will be back with a few changes at some point. no ETA.
Do you know any other paste service that works without JS (i.e. using curl) in the meanwhile?
Replying to
@SRAZKVT@tech.lgbt @DLC@mstdn.social the problem with CSS is not so much in design of CSS itself but in the complexity of HTML which make each page have its own set of unique CSS rules.
I agree with authors being able to suggest a set of styles, but those should be optional on the client. That way you can reliably set your own scheme for all the pages and still be sure that it will reliably work.
This is also important for readability, not only for the aesthetics of the page.
Replying to
@anthropy what about the irony of not being able to load the GitHub issue I opened about it without JS? 🥲
https://github.com/mastodon/mastodon/issues/34241
See also:
https://github.com/mastodon/mastodon/issues/6186
PS: I use bloat to access mastodon when I'm on my older computers, no JS required: https://git.freesoftwareextremist.com/bloat
There is also snac for no-JS server instance: https://codeberg.org/grunfink/snac2
Replying to
@gumnos@mastodon.bsd.cafe @olivetree@ieji.de @xinit@mastodon.coffee perhaps this could be a good starting point: https://alex.envs.net/
Replying to
@bohwaz thanks!
For HTML selection I normally use "dilloc dump | xq ...":
https://github.com/sibprogrammer/xq
Regarding screenshots, I already implemented it for rendering tests, but I need to test it a bit more specially under Wayland and other platforms.
For the scroll, I think you could just inject the Space key into window, which will scroll the configurable step size.
But yeah, it probably will grow to cover more actions in the following releases :)
Replying to
@SRAZKVT@tech.lgbt I don't have definitive answers to many of those questions.
Ideally it should be transport agnostic, so we can transparently replace http for another one.
Replying to
@zyd damm it:
Anonymous uploads are temporarily disabled due to agentic AI abuse and other various issues currently plaguing file hosts like Catbox.
Replying to
@emil in X11/Xorg there are two "clipboards":
The "primary selection": Selected text gets immediately copied, paste with middle button or Shift+Insert.
The "clipboard": Copy with Ctrl+C, paste with Ctrl+V. Many programs also store the content in the primary selection even if not selected.
I mostly use the primary selection because it is faster. Specially, middle clicking the [X] next to the URL, directly opens the URL in the primary selection.
https://wiki.archlinux.org/title/Clipboard#Selections
Replying to
@S-Config rough roadmap:
- Accelerate automatic testing so we can pass tests quickly (done).
- Get a baseline of WPT CSS2 tests, currently at around 20% or so, but they need to be manually verified (wip).
- Add the WPT tests to the CI.
- Fix the CSS CDATA and Ahem font support.
- Test some pending CSS fixes, so we can delimit regressions.
Note that CSS has less priority than being able to access web sites, which we have improved, but we cannot bypass all restrictions yet.
Replying to
@jwo yes, is the default search engine in Dillo. I hope they don't follow Google's path.
Replying to
@dheadshot @brewsterkahle yes, that classic frontend WM is great. After taking a quick look, they seem to have switched to web components that render in the shadow DOM. I was hoping this was some kind of bot-protection move, but it seems to be another case of enshittification (I hate that word!).
I hope I'm wrong and they still care about non-JS browsers. They still provide a search API, so building an alternative frontend seems feasible: https://archive.org/help/aboutsearch.htm
Replying to
@zbrando I'm afraid neither Gemini, Gopher or the "small web" would fix the current problems with LLM crawling bots or slop content, which could also surface there eventually.
What is a DOIs?
A "DOI" is an identifier used to locate research papers: https://en.wikipedia.org/wiki/Digital_object_identifier
Replying to
@jwo links is great. Yes, Dillo only works with FLTK.
The ideal scenario is that you can choose the browser that fit your needs, not that you are forced into a monopolistic browser market.
Replying to
Replying to
@SRAZKVT@tech.lgbt @navi@social.vlhl.dev @DLC@mstdn.social a bit similar to the Devine navigation system in XXIIVV, but without the need to request more pages to see the whole tree:
https://wiki.xxiivv.com/site/directory.html
Replying to
@SRAZKVT@tech.lgbt @navi@social.vlhl.dev @DLC@mstdn.social I was thinking if the navigation system could be built a bit different.
Like you press N and it would open a tree-like structure highlighting the current node which is the current page. Then you could navigate in a 2D hierarchy using only the arrows.
Pages wouldn't need to carry their own navigation links, but is already built in from the way you structure the site.
@c64whiz@oldbytes.space @rl_dane@polymaths.social I reported it back to Edegesuite (Akamai) last year (AKAM-CASE #F-CS-9320406), but they have not replied since then. Last time it was caused by the User-Agent header (http_user_agent) and changing it worked, but I haven't changed it now (Dillo/3.2.0) and it still works for me.
Replying to
@k@layer8.space @Foritus@toot.dusepo.co.uk @Li@tech.lgbt I would say it might work, as building Dillo with gcc only needs around 150MiB of RAM, and was done from my Kindle and old Android phone.
Also: https://www.youtube.com/watch?v=ORquozBaHBY
More examples here: https://dillo-browser.github.io/gallery/
Replying to
@S-Config Dillo has proxy support, but only HTTP not SOCKS. Notice that both HTTP and HTTPS traffic is sent via the HTTP proxy:
https://dillo-browser.org/user_help.html#proxy