2). or to port new-last webkit core and start to made new gui on top of it. Yes, crap about gui, but webkit core is more important. If you have at least few gui buttons to navigate and download manager, and nothing else, but webkit core are of today standards, it mean it can be used. In compare with full-blown gui, and 7 years old core.
Yep the second one of course, maybe later you can step-by-step reuse the current mui code with not much modifications so to "recreate" the mui-gui we have already in 1.23 ?
They work fine on what they need without anyone from os4 camp being connected with. They also do not want to be connected with aos4 at all and with all that Hyperion's mambo-Jambo, all issues, legality boredom and with Ben itself because of past. I also remember reading on morphzone that it was told cleary about new browser: it will be only and only for morphos.
Pity. A web-browser is one item where pooling resources would really help.
I've also been thinking about how to get a better web-browser. Like LiveForIt, just getting Odyssey/Webkit to build is too much for me, and it seems to be a moving target (as in, the WebKit team are doing large internal changes all the time and the endianness issues are likely to get worse). Google Chrome's build system looks even more complex, as-is Firefox I don't see NetSurf getting up to standard, as progress there is too slow.
An alternative option would be to create a new browser starting with litehtml for handling HTML & CSS, and Fabrice Bellard's QuickJS for Javascript. The render backend would be OpenGL or Warp3D Nova directly.**
EDIT: The build system would be makefiles only, and there would be a browser core (HTML, CSS, GUI, etc.), with everything else being plugins (e.g., video playback, maybe Javascript so the Javascript engine could easily be swapped out later with one that supports JIT).
LiteHTML's CSS support is incomplete, and its render engine doesn't handle dynamic elements, but it would be a good starting point. Plus, we'd need a GPU accelerated text rendering lib.
The biggest advantage of going this route, is that we could set up the code to have all dependencies present, and building would be a simple make. That way more people would be able to work on the code (or at least tinker with it).
It's all doable theoretically, but I simply don't have the bandwidth for it. The only way I could work on something like this is if I were fully paid, and nobody wants to pay for a webbrowser these days. Alternative funding options such as getting search engines to pay a revenue share (e.g., DuckDuckGo offers that) would require a large userbase to pay for development, so that's not viable either.
Just throwing this idea out there in case someone is crazy enough to try. Or perhaps it'll spark a better idea.
Hans
** OpenGL would be needed to make it portable so that the browser could have a large userbase.
It is not true that no one wants to pay for the browser. See, even archaic IBrowse is commercial. Earlier, no one gave us anything for free, the Timberwolf and Odyssey donated 7,000 euros. Only the old OWB and Netsurf are actually free.
I do not know what are the moods among amigos in the world, but discussing with amigos in Poland we have long come to the conclusion that an internet browser is such a very important program that it is worth paying for it. One could pay one bigger amount on start and then some smaller for upgrades, so as to provide constant motivation for the author to develop further.
An alternative option would be to create a new browser starting with litehtml for handling HTML & CSS, and Fabrice Bellard's QuickJS for Javascript. The render backend would be OpenGL or Warp3D Nova directly.
That alternative option mean to go the same way as many other tried : create a browser almost from scratch. That is unpossible for one person today.
Besides, while there is no DMA support with opengl/warp3dnova, that all out of question. It will hits the bounds all the time and limit usability.
Quote:
EDIT: The build system would be makefiles only, and there would be a browser core (HTML, CSS, GUI, etc.), with everything else being plugins (e.g., video playback, maybe Javascript so the Javascript engine could easily be swapped out later with one that supports JIT).
Today buld system really not important. Any big things will be build anyway on cross-compiler in favor of speed. Imagine that if i can build webkit on modern PC for a half of a hour, and on x5000 it will take about 10 hours. That unnaceptable.
So in any case, cross-compiler must, but if it must, then be it make, or cmake, does not matter.
Quote:
LiteHTML's CSS support is incomplete, and its render engine doesn't handle dynamic elements, but it would be a good starting point.
Its even incomplete ! So twice unpossible. Even if one will get paid fully, its just unpossible.
Why for example IBRowse still didn't get that route ?
Quote:
The only way I could work on something like this is if I were fully paid
It will take years, and will be not close to webkit internals. But all valuable time which can be spend on imporving opengl/warp3d/drivers will be lost.
Can you imagine follow always changed web-standarts ? In webkit there many developers, and they somehow cope with. Some of them paid fully, etc. You can't compare it with litehtml/quickjs, etc. Which even incomplete.
@mufa Quote:
It is not true that no one wants to pay for the browser. See, even archaic IBrowse is commercial. Earlier, no one gave us anything for free, the Timberwolf and Odyssey donated 7,000 euros. Only the old OWB and Netsurf are actually free.
"pay for browser", mean not 7000 eur one time. It mean real-world payment for developer work, it mean 2000-3000 each month for few years :)
7000 for ready browser its nothing, if we take out "we love amiga and do it for amiga". It's was only bonus for them. Its just 2-3 months of works of real world programmer on usual solary.
That alternative option mean to go the same way as many other tried : create a browser almost from scratch. That is unpossible for one person today.
Nope. Creating a browser from scratch would be to start from no code, write my own HTML/CSS parser, rendering engine, etc. I'm talking about using existing libraries and working with their developers. The libraries are selected as ones that should be easy to port & use.
Quote:
Besides, while there is no DMA support with opengl/warp3dnova, that all out of question. It will hits the bounds all the time and limit usability.
I don't think so. AFAIK, Odyssey, OWB, etc., all do rendering in software, so using a GPU assist should be a step up with or without DMA.**
Quote:
Today buld system really not important.
It is if the people who would like to work on it can't even get started. LiveForIt isn't the only developer who tried and gave up...
You don't seem to realize just how big a stumbling block this is.
Quote:
"pay for browser", mean not 7000 eur one time. It mean real-world payment for developer work, it mean 2000-3000 each month for few years :)
Yes, I'm talking about making a browser project financially viable, so enough income to pay for a small team to work on it. This would include paying the litehtml author to improve litehtml.
While I'm sure AmigaOS users would pay for it, we really need a large userbase to make it possible, and I can't think of a way to create a browser that Windows/Mac users would also pay for vs just using the free options already available. This is the true stumbling block to make such a project viable.
Hans
** Provided that everything is GPU rendered. Otherwise time spent shovelling data between VRAM and RAM would kill any benefits.
You don't seem to realize just how big a stumbling block this is.
Cmake the same easy as Make. It's just it cant be used on amigaos4 natively for real (as it containt some bugs), so cross-compiler must. For some simpler things make is better for sure (as can be used natively too, like SDL builds , etc). But in case with odyssey Cmake already there.
Its just was me who didn't wrote some instructions about odyssey's build exactly, but its not related to issues with Cmake of course. And in case with LiveForIt i just wasn't in interest, because after how he handling Mplayer by creating another fork "liveforit-mplayer" , instead of merging everything, i was't in big interest of "liveforit-odyssey" :)
When we do first attempts to port odyssey, i even made build of main libraries over cmake on crosscompiler, and give those libraries to Daniel from onyxsoft so he can compile amiga-specific code + link those libs with over make on his x1000.
Quote:
Yes, I'm talking about making a browser project financially viable, so enough income to pay for a small team to work on it. This would include paying the litehtml author to improve litehtml.
And that all mean that if there will be no one willing spend a lot of money, that route you think of out of question probably.
But then, it can be also easy to just pay to some developer in webkit who will just "support powerpc32" , and that all :) Webkit rendering over WinCairo its only one of renderers, it also can render today over opengl and whatever else. Back in past there were SDL, plain Cairo and co.
Our situation with no money and no userbase only leave us with webkit. Imho.
I don't think so. I've tried using CMake, and it's horrible once you start doing even slightly complicated things. I have instructions saved somewhere on how to use it with a cross-compiler, because I can't figure it out from the docs.
Quote:
And that all mean that if there will be no one willing spend a lot of money, that route you think of out of question probably.
As it stand, yes. I'm throwing the idea out there, because sometimes kicking ideas around between people can result in something that's actually workable.
Quote:
Our situation with no money and no userbase only leave us with webkit. Imho.
Maybe. I hope that we do get an updated WebKit working complete with WebGL, video playback, etc. I do think it's worth examining other options, though.
NetSurf might be wort looking at, it sable browser that seams that can be good foundation to build on. better then start from zero.
Already looked at that. As I said in my original post, progress on that one seems too slow. I also don't think that writing a browser in pure C is a good idea; you're just making things harder for yourself.
Likewise, I looked at using some of its components, but LiteHTML and QuickJS just looked like better options. Using existing HTML/CSS and JS engines is far from starting from zero...
I don't think so. AFAIK, Odyssey, OWB, etc., all do rendering in software, so using a GPU assist should be a step up with or without DMA.**
There is few examples when things rendered via software (or better compositing) _much_ faster when they rendered via opengl (does not matter minigl , or gl4es). Without DMA support it not worth to even experiment with : once you will need to update texture per frame, it will hits the bound of slow texture updating because of no dma and hitting 100% cpu.
Examples which prove that SDL2 and DosBOX : it just show that uploading of textures are slow and CPU is busy with that, and things which should be faster, start to be only slower because of no DMA. Same for all big games : loading of textures just take time and cpu busy with that.
By some reassons uploading textures via opengl/nova are pretty slow. And reassons probably RAM<->VRAM transfers which mean DMA is must.
Imagine someone start to made a ground-hard work on it, and later realise no-dma issues, and then hold project for years because dma can not happens by any of reassons.
... Without DMA support it not worth to even experiment with ...
1. Such a defeatist attitude won't get us anywhere 2. Check the footnote that goes with the bit you quoted. I'm well aware of situations where RAM<=>VRAM transfers would outweigh the performance gains of GPU rendering, and I do have some idea of what's required for rendering HTML/CSS 3. Lack of DMA/GART won't last forever
Most of the web technology needs to be abstract away into system wide libraries, so all windows have web tech. Why can’t my window be broadcast to my screen? Why can’t today every software be web aware so even a paint package could have own chat, broadcast a stream or get plugins; see and share users work, inter operate web imaging api etc all via system wide web datatypes. If Hyperion had vision we would have these libraries and tools to help create better software quicker.
Realistically issues now is : 1) or to port fresher odyssye from aros repo (so we will have the same odyssey, just webkit core +2 years in compare with our) 2). or to port new-last webkit core and start to made new gui on top of it. Yes, crap about gui, but webkit core is more important. If you have at least few gui buttons to navigate and download manager, and nothing else, but webkit core are of today standards, it mean it can be used. In compare with full-blown gui, and 7 years old core.
If the first implementation is a matter of 1-2 weeks, then I would say, go for it. This will buy some time, maybe a couple of years, until the new browser is implemented.
But of 'course the second solution is the best of all. Have you thought that a good GUI might be IBrowse? Maybe IBrowse team is interested for a good webkit engine for IBrowse 3, and would like to contribute and work on that "Super Duper" browser.
I am sure that if people paid 30-60€ for IBrowse 2.5, would be glad to pay the same amount of money for a much more capable browser. But, unfortunately, I don't know how many will do that. But I am sure the amount of money will be difficult to cover the basic salary of one developer. But a bonus is a bonus.
@Hans My opinion is that webkit is a much better supported and updated engine, and if it is possible to be ported, that the way we should go.
On the other hand, instead of creating a new browser, I also agree with LiveForIt, that it would be better to work with NetSurf team and make JS even better for that browser, since HTML and CSS is in pretty good stage. I understand that it is developed in plain C, but I am not expert enough to understand if it is not enough for a good JS engine. At least with NetSurf you don't start from zero.
Now, if this should work over Cairo or OpenGL/Nova, I would prefer the later, since this is more updated right now, thank to the work Hans, Daniel, Roman and ptiseb do. But I don't know the technical details and I can't be sure on that, but I see it more easy to reach a GPU accelerated engine than with Cairo.
At the end, unfortunately, my knowledge on programming for the above stuff is so limited to nothing. I wish I could code for such a project. Maybe I could do something in the GUI side, but not too advanced things. But I could help on creating a stable, consistent development environment for the team that is going to work on that browser, based on a docker image, for cross compiling. I have experience on that and the benefits can be a lot. And of 'course I will be one of the first persons to put my pocket where my mouth is, and support this project.
... But I could help on creating a stable, consistent development environment for the team that is going to work on that browser, based on a docker image, for cross compiling. I have experience on that and the benefits can be a lot. And of 'course I will be one of the first persons to put my pocket where my mouth is, and support this project.
If you could create a docker image that sets up the entire dev environment including all dependencies, then that would be a great help with updating Odyssey & WebKit. That would drop the time wasted collecting dependencies, setting up the cross-compiler and figuring out how to even build the thing down to pretty much zero.
We seem to have mostly two kinds of people: those like kas1e who are good at porting (incl. dealing with CMake, autoconf, *nix dependencies), and those like me who are good at writing software. Very few people are good at both. With setting up the dev environment out of the way, then more developers would be able to work on the code.
@hans I would be glad to help on a team that will start working in such a project. And you are right. Creating a consistent dev environment which is easy to install, no matter the OS anyone uses, with clear documentation on how to do it, is a very good solution for anyone to start working on something.
Are you guys interested on forming a team that will start to organize the whole implementation, build/find the necessary tools, and do the investigation for a valid solution, before anyone start the development?
I checked yesterday deadw00d's github, where Roman's branch has his own ported code, and I found joakim's branch where he has a few changes in the code, a helpful guide how to compile and a dockerfile that is for compiling Odyssey. I haven't test it yet, but there is something for that.
News feed
OS4 Feedback forum
OS4Depot Feedback forum
Amigabounty forum
OpenAmiga forum
AmiCygnix forum
ABC shell forum
AmiKit forum
Cinnamon Writer forum
CodeBench forum
Digital Universe forum
Dopus 5 forum
E-UAE forum
Gnash forum
Ibrowse forum
JAmiga forum
Odyssey forum
OWB forum
Qt forum
SmartFileSystem forum
Timberwolf forum
TouchDevice forum
TuneNet forum
Unsatisfactory Software forum
Donate
