Hacker News Reader

Comment by smallstepforman | Hacker News Reader

Here is a representative screenshot of Haiku environment on 4k (3840x2160) screen running my native Media editor for Haiku (audio+video). It took me 2 years of a couple of hours per week to build it from scratch.

https://raw.githubusercontent.com/smallstepforman/Medo/main/...

View on HN

That’s very impressive and very cool!

Maybe you can say something about this, given your in-depth work Haiku and media:

BeOS was famous for being able to run all sorts of media content on relatively slow hardware, that any other system would struggle with.

Is that still true for Haiku? Or did the other OS’s catch up, or maybe the hardware did?

Haiku relies on ffmpeg to decode videos, so there is no difference between other systems. BeOS used to have yuv420p, yuv444 etc supported colour spaces with supported drivers, and the interface kit BBitmap supported YUV colour spaces via overlays, so you gain efficiency by doing one conversion less. Haiku lacks proper drivers so has to pay the price. For a video editor which needs BGRA colour space, you’d use shaders (SpirV or GLSL) to convert colour spaces. Haiku lacks HW accelerated drivers so it loses there as well (compared to mainstream OS’s). Also, Haiku has no vaapi HW driver so it loses there as well.

On the plus side, Haiku has a fantastic API and with BMediaKit and translators, decoding a video stream is a couple of lines of code (much nicer than raw avformat API from ffmpeg). The async bitmap transfers/uploads are multithreaded so it wins there. Haiku allows efficient sharing of media buffers between apps, so drag/drop between apps is really efficient.

I went overboard with making the video editor as multithreaded as possible. A thread for thumbnails, decoding, audio mixing, opengl, the main gui window, preview window, effects gui, encoding, etc. They’re all threads. The editor is really smooth, and can do 4k editing. I cant wait to see how it will runone day with actual HW accelerated drivers.

IMHO every fringe OS community should invest effort into developing good full-fledged drivers for at least one reference hardware configuration. I wouldn't mind buying a specific video card, a specific motherboard etc even if they were rather old and overpriced in case if Haiku had stable hardware-accelerated drivers supporting all their features.

Given open-source Linux drivers which can give a good clue about how to interact with specific hardware this task doesn't seem impossible.

OT/ I hardly ever see “fringe OS’s” port Linux drivers, but a lot of them do port drivers from NetBSD. I wonder why?

That's explained here: https://news.ycombinator.com/item?id=28218619

NetBSD drivers are particularly convenient to port, also because of the license.

I don't think it is reasonably easy to port Linux drivers but they can be a good source of inspiration and hints when engineering drivers of your own.

This looks like a great app! I'll have to check it out. I wrote a video editor for BeOS back in 1998 called UltraDV. Nothing really happened with it, but a lot of the code continued to live on and evolve in the Muse Reasearch Receptor hardware VST box.

Thanks for the feedback. I never tried UltraDV, however I did have a copy of personalStudio.

Regarding VST plugins, I do have the book "Designing Audio Effect Plugins in C++" by Will C. Pirkle and plan to add VST plugin support to Medo. However, I know nothing about VST use cases so I'm developing in the dark here ...

It’s cool knowing that this was developed in Melbourne. I still feel like it’s a little city in little Australia, but I know there’s some crazy talent in town. Thanks for you hard work!

Regarding VSTs in a video editor: for doing any dialogue work (eq, compression, de essing) would be some use cases I have (I use some VSTs in FCPx). Also for working with music (same eq, compression, gating/ducking). Not having to bounce the audio through a DAW of some sort to just get it to sound right is a huge productivity boost.

That said I am always surprised when I see local stuff that is of high caliber. It always feels like it is a little guys proving it on the world stage.

That said... Little city!? I surprised that folks would see this place as small. I would have thought that back in the 90's but we have slowly become, maybe a little too much, a big city.

For anyone curious, Melbourne has 5,159,211 people, which puts it ahead of fan-favorites in the US like Los Angeles, Chicago, Philadelphia, San Francisco, Boston... and literally every city in the United States with the exception of New York!

In fact, it has more people than the states of Wyoming, Vermont, Alaska, North Dakota, South Dakota, Delaware, Rhode Island, Montana, Maine, New Hampshire, Hawaii, Idaho, West Virginia, New Mexico, Kansas, Mississippi, Arkansas, Nevada, Iowa, Utah, Connecticut, Oklahoma, Oregon, Kentucky, Louisiana, Alabama and South Carolina.

Oh, it also has more people than Puerto Rico.

> For anyone curious, Melbourne has 5,159,211 people, which puts it ahead of fan-favorites in the US like Los Angeles, Chicago, Philadelphia, San Francisco, Boston... and literally every city in the United States with the exception of New York!

I tried to look up what the "metro" population of Melbourne is, but could only find a reference to "greater Melbourne" being around 5M. LA metro area is 12M.

Is Melbourne really larger?

Don't deceive yourself. The numbers involved and the underlying reality are a little different than what you're thinking, since it all depends on where you draw the lines. The entire state of Victoria, including Melbourne, has something like 6.7 million people. [0] The United States has 11 combined statistical areas [1] with populations greater than Victoria. The way large cities are laid out, it often makes little sense not to include the suburbs.

LA County and the Chicago metropolitan area each have roughly ten million people, and NY metro is at least twice that. They are really, really big. The most interesting thing about Melbourne and Victoria is its rapid population growth. The graph might not be a hockey stick, but it is extremely steep.

[0] https://en.wikipedia.org/wiki/Victoria_(Australia)

[1] https://en.wikipedia.org/wiki/Combined_statistical_area

Major props to you, this is really one of the best post-BeOS apps available on Haiku.

Thank you, it was a labour of love :) I still have an enormous amount of work to do, eg. adding many more audio filters and effects, text effects, programable OpenGL effects, and of course bug fixing. Hopefully in a decade or so it will be a solid dependable application.

Looks impressive! Congrats, but more than the software I admire your persistence on a side project over two years.

That's a really great result, nice work.

I'm curious what it's like to develop media applications for Haiku. Are there any advantages you felt made it easier or more enjoyable?

Thanks for sharing.

I enjoy the BeAPI. In my opinion, the Interface Kit (GUI) has just the right balance between having enough features to be useful and not getting in your way. The BMediaKit realistically has 2 purposes, decoding/encoding and streaming and ideally should be split up into 2 kits. The Haiku based packaging system is really great for developers, since installing headers/libs requires zero effort, it just works. Tracker (file navigation) is fantastic, finding header files takes less than a second. Deskbar needs to apply 20 years of modernisation and adopt features from other OS's. The browser situation is terrible, though. System administration is really well thought out, however, it needs user accounts and encrypted volumes.

Regarding other development environments, I've done professional Qt development, and I'm displeased with the ownership and kitchen sink model. Ownership and auto layout make dynamic widget hiding a real pain, especially since you cannot remove a widget (ownership issues). Layouts with dynamic widgets are another pain point.

> it needs user accounts

Why? It's a desktop OS. The use case has a median user count of 1.

A family PC has more than one user. I'm sure that the same people don't want to, say, share their browsing history and log-ins or have the same taste in wallpapers.

You could do an n-way multi-boot, with an instance of the OS for each person. It's definitely small enough to support that.

> A family PC has more than one user.

Do these really still exist[0]? As far as I can tell, they only ever existed during the window between when internet access became ubiquitous and when smartphones became ubiquitous and laptops got really cheap.

Besides which, I think there are better ways to accomplish that goal without user accounts. User accounts exist to protect the system and users from other users in a shared system or network, which doesn't really apply in a Desktop use case. Instead I would say what you want is switchable user profiles, plus encryption to actually protect your documents if that's important to you, because it is trivial to bypass filesystem permissions for local users anyway.

[0] I'm sure somebody is doing it, but I have to imagine that within the vanishingly small Desktop PC market this use case is itself vanishingly small.

Reply on news.ycombinator.com