Started by Mark bc7291552be7a58f... ·
Oh! A wild image appeared! Just noticed the screenshot in the first post of this thread now loading as an actual image. Nice work @Zenith! Didn't expect it so quickly ;) Was a bit weary of adding extra work for you with all of this ;)
Good to hear @K8 and @burger! Was especially interested in whether it would actually work on macOS @K8, as I didn't have a mac to try it out on ;)
This is great Mark, thank you! Images on Nomadnet was precisely what I needed: I'm already in the process of trying to mirror my internet site to the Reticulum network. Don't worry to all the pixel haters: my static website is already mostly text based and fast loading. For the Reticulum mirror I will keep it that way.
Thanks Mees :) I really enjoyed reading your site and seeing the photos of your networking gear! And your antenna mast base with an actual tile roof absolutely made my day, that is a thing of beauty :D Thanks for sharing it! Really cool that you're uplinking it with the 5GHz PTP link too.
Mark wrote:
Regarding progressive / bandwidth selective loading, it may actually be simpler and more optimal overall to allow sending a
profilekey on the request (similar to LXST profiles), which will dynamically downscale and increase compression level at the node, before sending (that was my initial plan).
I love this idea! Could the same profile key also cover Resource compression? bz2 needs the whole block in RAM before it emits anything, so on a 256 KB MCU (nRF52840, microReticulum) pages cap out around 24 KB decompressed and can't stream. DEFLATE handles any page size in a 32 KB window and streams.
Something like "deflate instead of bz2", and the node serves the page that way. Nodes that don't know the key do what they do now; the client falls back to uncompressedd Resources, which already work at any size. Either nomadnet deflates the page bytes and hands them to Resource with auto_compress off, or Resource gets a codec flag. Would either fit the profile system as you have it?
One of my favorite websites uses dithered images by default to preserve bandwidth. You all are probably way ahead of this but just wanted to throw the idea out in case it's useful.
b1LL wrote:
One of my favorite websites uses dithered images by default to preserve bandwidth. You all are probably way ahead of this but just wanted to throw the idea out in case it's useful.
Or more accurately, the website serves dithered images by default with the option to view full images if desired. I appreciate having the choice.
It's a really cool website, but the dithering is silly (looks really cool, though). For the "solar powered foot stool" article, the main image is 26 KB as their dithered PNG, monochrome. But a full-color version, with more visual information (just slightly below the threshold where compression becomes perceptible) is only 9.6 KB, so 2.7 times smaller!
So if they actually want to save bandwidth, they could just use a better webp encoding preset :)
hi, i'm currently using a mix of colorado mesh client and nomadnet navigator and just wanted to say this looks really cool!
Zenith wrote:
If you want a really fascinating rabbit hole, take a look at the art people have made with JXL's predictive trees
https://jxl-art.lucaversari.it/wtf.html
https://jpegxl.info/art/Welp time to add images over in Nomad for this forum! You should see that soon
This is indeed quite the rabbit hole! Love the demoscene angle.
