learned today that people *really* underestimate how complex digital video signal transport is
i examined my assumptions and learned that they include "ah, I think everyone probably has at least a basic understanding of gigabit line codes and SerDes". that... that would explain why
@whitequark@mastodon.social it’s just taming some electrons, Catherine, how hard could it be
@aud @whitequark it's one gigabit/sec michael, what could it cost, 10 dollars?
@aud @whitequark it’s just engraving some* runes and imbuing rocks with a soul Nancy, jeez, we’ve been doing that since we were in caves, how different could it be
*a few hundred billion
@whitequark I feel like something about how prevelant / common / "easy" it is also probably plays into this quite significantly.
@attie counterpoint: I think everyone has tried to connect their laptop to a projector at least once
@whitequark ... my brain has now gone down the rabbit hole of people's general understanding of video compression and general data transport too.
@fclc I was referencing that!
@whitequark reminds me of a presentation I gave where I assumed everyone could quickly read and parse nested grouping regex.
I was giving a talk to Ruby programmers in the lovely transition from “Perl programmers who think functional programming or japanese culture are neat” to “Rails web programmers”.
I think I spent 20 min answering questions about slide 3, then just gave up on the plan.
Sounds like an extremely interesting thing to learn about, but at the same time, I definitely think I have a ways to go on the foundations to be ready for that.
@whitequark I was looking at capturing video from DSI output on a device and looking at all the terminology just threw me into a "holy shit this is actually way more than I thought it was"
@whitequark there's an xkcd for that!
https://xkcd.com/2501/
@whitequark Slightly related: I was pretty surprised to discover that both VGA and HDMI have I2C lines.
@oscherler "we have two spare pins and a need to record 256 bytes of data for EDID!" "do you know what uses two pins and can store 256 bytes of data? that's right, 24C00!"
@whitequark DP really isn't shy on complexity, with its link training, multipath routing and countless other features
@whitequark Honestly if I were designing a digital video protocol in 2024 I'd leverage as much existing open standard tech and investment as possible while maximizing interoperability and flexibility.
The standard connectors for consumer applications would be RJ45, SFP+, and LC, with QSFP+ and MPO as higher-performance options targeting professional/studio end users. DACs being used for short range and fiber / EOE for long range.
Framing would be IPv6 over 10/25/40/100G Ethernet, using link local addresses by default but with routability possible for more complex topologies.
Replace EDID with a standard discovery protocol, either HTTP to download a metadata file from a well known URL or maybe something mDNS based that would simultaneously provide device discovery.
Video transport would be uncompressed RGB24 (maybe with YUV and high-bit-depth as optional addons) over UDP, probably using RTSP or similar.
@whitequark And this is why more and more protocols keep getting heaped on top ... once you're over that phy layer hurdle of establishing some reliable bit-lanes ... why not?
