I promise, this isn't a troll. In fact, I'm not even an Amiga user, but
I can guess some of the answers to the questions.

I've had a big of a dig around, and of course I've seen the AmiZilla
bounty, but why is it taking so long?

But before you answer, some background. I'm the guy currently porting Firefox
to RISC OS. That's the OS that was developed to run on machines which were
32-bit developments of the original BBC Micro which of course competed (at least, in the UK) with C=64s. But that's ancient history of course. If you're in the UK you should be familiar with Acorn and RiscPCs.

Browser development on RISC OS has been characterised by many commerical browsers (around 6) which for the most part, stopped when the company
in question lost interest. Lots of versions of similar things, none of
which are 100% satisfactory is a sad part of the history of RISC OS browsers.
I presume something similar might be said about the 3 main Amiga browsers.

I even wrote a page about this issue, which is the basis for my talks at RISC OS shows: http://www.riscos.info/unix/browser.html

And now, thanks to enourmous efforts by a very small number of developers (probably far far fewer than the Amiga ones), various Unix compatibility has been improved (especially the C library), porting tools have been developed and sheer determination, I'll hopefully be releasing the first beta of Firefox for RISC OS in the next few days. It's even just about usable on the 11 year old RiscPC hardware that many RISC OS users still have (newer hardware is available). A Miracle, no less. More details here:

http://www.riscos.info/unix/firefox/

The process has been long, and expensive. In many ways, RISC OS is probably less suitable for running Firefox than AmigaOS is: no PMT, no shared library system until probably much later this year, fewer developers.

I appreciate that the current AmigaOne stuff may well have been a distraction from getting much done, as well as all the other headaches of a small platform.

So, my question again: Can anyone give me a (sordid or otherwise) history of Amiga web browser development, and explain why there isn't a Firefox port yet.

Thanks.

@Chocky

There isn't a port because nobody has ported it yet. It's that simple.

@Chocky

I guess its because we have more committed users than committed developers. We have a split community with in-fighting. We have no clear choice on what gui-toolkit to use. We have no port of GTK. We have no time. We have plenty of excuses

_________________

@Chocky

Probably because there are not enough capable coders. Most of them work on some next generation AmigaOS system, like AmigaOS 4, MorphOS and AROS.

And, of course...AmigaOS is not Unix. This makes the basic toolkits hard to port.

@Chocky

I've often been wondering this. Back in the mid-90s and on, just about any OpenSource app available for Un*x had a port to Amiga (via the excellent Ixemul.library, which allows Un*x apps to be ported to Amiga). Some of these apps didn't run in the Amiga's native GUI, but on an X Server started on the Amiga! That got around the graphics issues, but some ports were given new GUIs. Command-line ports just ran from the Amiga shell.

Now I guess it's because Mozilla/Firefox uses newer luxuries such as threads that it has not been swiftly ported, even to X11 just to get it going.

Capable compilers are still available so they're not an issue.

I'd be interested in what sort of rendering speeds you get on RiscOS. The main workstation in use at home runs Linux 2.4 on an AMD K6-2 @500Mhz, which should be plenty; however it crawls! So I'd be interested to see what you get on a 200Mhz Risc Machine. Is it useable? Maybe it's X11 making the rendering so slow on Un*x...

@Chocky

I can try to answer your question.

The FireFox port for Amiga aims at supporting not only AmigaOne, But even MorhpOS and the
classics with OS3.9. And none of thema is truly UNIX. (remarks: i mainly speek for AmigaOne
and classics, because my knowledge at MorphOS is limited). The main problem lies in establish
a build env. for FireFox. On the classic there are the most needed tools available, but sometime
not the needed version or sometimes they need a patch. The setup is actually quit complicated.
There are undertakeings to simplify the process, but it goes slowly. At the moment there is guide
written for setup a cross compiler trageting AmigaOS 3.9.

For AmigaOne and AmiagOS 4 not all needed tools are yet available. But some guys seems to
port them and from the OpenOffice port seems to come out alot of tools needed by FireFox and
OpenOffice.

A good news is that one developer have ported the nsrp part for AmigaOS3.9 and it worked, but is
still have had luck to reproduce it. Mainly because my time is limited to hunts bugs in my setup.

Resume: There is still a high border to overcome before one can get active in the development.
Thus there isn't yet a working port.

I wish you the best for your porting effort.

@MigthyMax and others

Allow me to try and answer some of your questions.

Firstly, the name of the OS is "RISC OS". I say that, because names are important.


Excuses - well, yes Excuses are a good starting point for finding out the kinds of problems that need to be solved.

GTK and toolkits. Yes, obviously GTK would a good thing to port - a great number of graphical applications that are interesting to port are GTK. I have a GTK 1.2 port; the GTK 2 port still requires some more work to detach it from freetype. They both talk directly to X11 - more on that in a moment.

As d4m0n has suggested, the is sometimes a choice between doing a native port using X11 and writing a whole new GUI backend for your OS. The latter is time consuming, and for many applications, it's entirely impractical. Both these are major considerations when you have so few developers. If the port uses GTK 2, then there's a possibility of customising the GTK backend, so that everything behaves much more like a native application. Plus, this is a one off effort that benefits lots of programs.

Few developers. I'm willing to bet RISC OS has less. There's probably less than 10 developers spending significant amouns of time outside of a couple of companies. There are lots of novice programmers, but most of these work on pet projects in their spare time. A good deal of my own effort has been to improve toolchains and libraries and documentation in order to make it much eaiser for all these groups.

"AmigaOS is not Unix". Indeed, but I bet RISC OS is even further away. That's why we have invested a huge amount of effort into the Unix compatibility in our C library. This represents about 10 years' effort, although the efforts of a couple of people in the last 2-3 years have made it really useful. This library is as near as to POSIX as is practical, and there are still bugs and misfeatures that are discovered every month. We have taken lots of code from glibc and elsewhere, and it includes pthreads and pretty much everything else you'd find in a standard Unix C library. Still, there are lots of things which are far from ideal, since RISC OS has very differing concepts about some things. In time, I see us having a glibc port, but that is perhaps 12 months off.

Tools: we use GCC 3.4.4. To be fair, I must point out that there's a commerical compiler which is very good and comes with lots of useful tools, but versions of GCC in the last couple of years produce significantly faster code, have full C++ implementations, and of course, are open source. The commerical compiler is faster to compile stuff (natively, of course) and until just a few months ago, did one thing which we'd wanted GCC to do for about 6 years. GCC now does that thing (just about, anyway).

GCC 4.0 will be ready soon, and with that we'll be transitioning free software development to ELF, instead of the native RISC OS object formats. This makes shared libraries easier - a desperately needed feature.

Build enviroments - almost all of my work is with cross compiling. There really is no other way with complex configure scripts and build systems. RISC OS is a long way from having the right kinds of tools to be able to handle anything but simplistic Makefile build setups. I'm sure one day it will be possible to do this, but for now, building under x86 Linux, on a much faster machine, where it is possible to actually run the build system - usually without any modification, is much more appealing and a huge time saver. I've also developed a range of porting tools that are wrappers for things like GNU configure and make which do all kinds of clever things and pass standard options. These too, save lots of tedium and potential for error.

X11: returning to this issue. RISC OS does have X servers, but none of them are very satisfactory, nor are any rootless (although I do have a port of Xnest in progress). Because of this, I took the mascohistic view that it would be "better" to instead replace much of the X11 library, and instead write a version for RISC OS that talked directly to the OS, and therefore served as libX11, XServer and Window Manager all in one. With 400 functions, this is clearly a lot of work, and there are lots of places to get things wrong, and performace to suffer if you don't get it right. But now, nearly 3 years later, this is nearly complete, and my efforts are starting to pay off.

What this means in priciple is that any X11 based application (whether it using X11 directly or via GTK or some other toolkit) can be linked with my library and then run directly as a RISC OS application. Sure, you potentially lose some performance and use a fair bit more memory than a native app - but it means the port actually becomes possible at all. I'm not sure if I could recommend a solution like this for AmigaOS - it is a lot of work. Perhaps improving your XServer would be a better option, but I think the advantages are obvious.

Returning to Firefox: The build I'm currently doing is of course cross compiled. The nspr port I have isn't perfect, but seems sufficient. The produced binary is a massive static 17MB. I am using the X11 build instead of the GTK1.2 build, since GTK still has some performance issues with my X11 replacement presently. The speed really isn't great on 200-300Mhz StrongARM machines. This isn't a surprise with such old hardware with oft-mentioned slow bus in a RiscPC. But I do expect it to end up being a little faster than the most featureful commerical browser. Of course, modern browsers do so much more than the older ones, so slowness is perhaps expected. On the newer 600Mhz XScale machines, the performance is much more acceptable, but I am still looking at where the bottlenecks are.

The port will use the X11 drop down menus (all RISC OS menus are popup on the middle mouse button) and other things foreign to us such as file dialogues (file loading and saving is drag and drop). In time, these can be improved either by modifying GTK as I suggest above, or adding to the X11 backend in Firefox.

I there was one lesson I had for you guys (without teaching you to suck gnarly's lovely yellow lemon), it would be to think reusability and automation. I know that native apps will continue to play a curical role in your platform and are much of its appeal, but the bottom line is that ports are going to fill roles it's simply far too much effort for a small platform to pull off by itself. Therefore, think about ways to do things that will help lots of ports, not just one, and think about ways to make it easy for other developers to reproduce your work, such as the autobuilder I developed which means many RISC OS ports can have their source fetched and built and uploaded with just a single command.

Anyway, after that essay, I'm still interested in the history of browser development on AmigaOS, if someone could give me a run down.

@Chocky

Browser rundown: from what I can remember, first was a port of Mosaic, about 1994? This followed by the commercial products IBrowse, AWeb and Voyager, some of which (later on) could use plugins like Flash (all eventually had a javascript subset with varying compatability/capability).

Voyager had the most features, but I found very buggy (regular crashes); IBrowse was the "most commercial", and quite stable; AWeb looked very sparse but was fairly stable and looks more like an Amiga application (at least a classic one) and came with the two last versions of the OS.

Of the above, AWeb is now OpenSource, with two versions: the original "classic" version, and a new "next-generation" version being developed as a "current/next generation" browser which is a work-in-progress thing. AFAIK AWeb is the only one being continually developed (don't know about IBrowse on AmigaOne). See here.

A mozilla port would be a favourite in my opinion but then it would be too slow on the classis Amiga unless the rendering could be done on the PPC Blizzard/Cyberstorm cards (the M68K series would be a slug with regards to the gecko layout engine).

BTW I ended up having to put Opera on a NetBSD machine (100Mhz Pentium I) because firefox was taking over a minute to start (and just as long to render a page). Opera, although being commercial, was very useable on that machine (this was using the Linux package with Linux emulation on NetBSD - Opera don't do a direct port).

BTW I know very little about web-browsers other than the protocol HTTP!

IBrowse is heading towards 2.4 release, which should fix a huge amount of bugs, I am not sure about new features. Version 3.0 should be PPC-native, but that might be a long waiting time, especially when the authors have many other projects too. Currently IBrowse runs emulated on PPC environments.

Voyager apparently had a PPC-native version for MorphOS, but as one of its developer ended into a conflict with his team, its future may be unclear.

As said, AWeb is open source. Sometimes I hear rumors that people are integrating KHTML into it. Good luck to them too. Aweb has naturally native PPC builds.

Chocky, very interesting read. I have no idea how far the Amiga ports of Mozilla are by now. I have only heard that people are divided into teams and competing to win the promised money.

Also new browsers have been announced, as Paihia. Demos are not available yet.

EDIT: this is an interesting thread and even Amiga-related...please put on the front page ;)

Last edited by Mr_Capehill on 09-Jun-2005 at 03:14 PM.

@Mr_Capehill

For completeness and fairness, I forgot to mention another RISC OS browser,
called NetSurf (http://netsurf.sf.net). This browser too is significant to the story of browser development on both platforms. Not only is it very small and fast, but it also implements vastly more in terms of CSS, and other modern web standards than any other released RISC OS Browser (the next commerical browser release is perhaps a few months off, and Firefox is not ready for a few more days).

NetSurf is also open source, and there's a GTK build if you wanted to try it out on Linux. Its main failing is that it has no JavaScript, and probably won't do so for a long time.

But back to the development of it - it too uses the Unix C library, and it makes heavy use of existing Unix libraries ported with the aid of stuff I mentioned earlier (e.g. cURL, XML libraries). There has been a fair bit of cross over between NetSurf and my work on Firefox and other ported applications. This is why I stress the reusability issue. I'm sure NetSurf could even easily be ported to AmigaOS

@gnarly
I don't think RISC OS has GTK yet either, it sounds like GTK calls are translated into native window widget calls via 'ChoX11'?

@Chocky
You may want to enquire on the mozilla AOS porting dev list, or wait a bit, as I don't believe many of the porters visit (or post?) AW very frequently.

BTW, the
UNIX Porting Project is interesting (subscriptions for *nix ports)..how is that working for you, and is it a solo effort or multiple developers?

Either way, good work on getting FireFox ported!

_________________
Are we not done with the same silly arguments and flames yet??!

@wegster

Strictly speaking, GTK on RISC OS is a port - although it is strictly just a straight recompile - which of course is ideally what you want for most unix libraries. It just hasn't been customised in any way for RISC OS. It talks directly to X11 and therefore to ChoX11.

The UPP is going well enough, but hard to keep up with administration of subscriptions, which is really not what scarce developers want to be doing. I am presently trying to find someone to do this for me.

The UPP is mostly my effort, although there have been some important contributions - GCC/C library is 4 people including myself, ChoX11 is mostly myself, with one guy doing much of the graphics. It is hard to do as much work as I do on essentially free software - I need to pay bills. I won't say how much, but I have had to raise quite a large amount to justify doing the Firefox port - even then, I will just break even. I think the AmiZilla booty needs to be much larger, and have a single development team. It probably doesn't help that the PayPal link is broken.

@Chocky

There are loads and loads of competent AmigaOS developers, but as you pointed out the effort to create a Unix-compatibility library takes years of development, and one really needs a core group. My vague impression (possibly wrong) is that the core group for AmigaOS, being the Geek Gadgets crowd, created and maintained an excellent Unix compatibility library for years, but sort of evaporated away when they got real jobs and/or moved to other Unix-based platforms.

I could be wrong, though, and I hope I am.

_________________
I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me.

@simplex

Well, if the unix library is/was so good, surely it's a good idea for someone to look at continuing on the good work? After all, it forms the basis of so many potentially important ports.

But back to Firefox - you have a working nspr, an X Server, and presumably a working cross compiler setup. With this in place it should be (relatively) straight foward to procude a version of Firefox using the X11 interface (not GTK) - such a version would just about, or even fully, qualify for being a beta version. Like most RISC OS users, I'm guessing that a port of Firefox that worked at all would be very interesting, even if the interface was quite different from the native in any initial version. This can come later - the point is that using the existing X11 interface requires no extra effort, and the existence of any kind of beta would certainly spur on other developers to help.

I'm not trying to be patronising, or pretend I know about your platform of course (since I don't), just trying to ask a few questions to get you guys thinking.

@Chocky

Firstly, to AW.net.

You provide a very interesting read & some pretty valid suggestions. I don't know exactly what's going on behind the scenes with our developers on a long term basis (They do keep quite nicely updated as they build apps for OS4). It's possible that once OS4 Final is out that some of our resources will be concentrated in other areas.

Your ideas sound quite realistic. Our developers may even have similar plans to your suggestions already, who knows. Either way, these sound to me like some good pointers.

FuZion.

@Chocky
WeLcOmE to AW !

@FuZion

>Your ideas sound quite realistic. Our developers may even have similar plans to your suggestions already, who knows. Either way, these sound to me like some good pointers.<
There has been plenty of discussions about which browser to port to AmigaOS4, and even some little fights...
If I was a developer working on a new Browser or a port of an existing, I would more likely keep it to myself until it's finish not to be "flooded" with "You should do this" or "You shouldn't do that"... Everyone here seems to have their own idees about which browser to port to AOS4 and which features it needs to have, but noone agrees except on the fact that AOS4 needs a more up-to-date browser...

Now, unfortunatly, I am not much of a developer ... But who knows, maybe something will come out with AOS4 Final ?

Jerry

_________________
- AOS has been ported to ex-86 ! It's called AROS and WinUAE... Or E-UAE on Linux !

- A1XE-G4 up and runing with:
512MB Ram / 200GB and 80GB HardDisks on Sii0680.
AOS4 Final Update / AmiZilla 0.1 Alpha

@Chocky

The UNIX library (ixemul.library) is good for quick ports, but it has a number of problems. Of course, there is the inconsistent interface (UNIX file paths and the like), there is a problem with vfork() (or is it fork()? I can never remember) not being possible on AmigaOS, and finally ixemul.library does not work on OS4. Actually, it does work, but the working version isn't available to us mere mortals yet. OS4 is more UNIX-friendly anyway, and ports which needed ixemul.library previously can be compiled without it using OS4 gcc.

Your idea of compiling just with X11 is a good one. However, the X11 servers for AmigaOS I understand are not very good (I've not had cause to use them myself). The best one is shareware, and the developer has gone AWOL. An X11 library like yours would be a treat.

There is (or was) a dodgy huge early compiled binary of Mozilla, compiled with ixemul.library and no graphical output (maybe it worked with an X11 server, I don't know). I'd be perfectly happy with a non-Amigalike port initially. We have a port of FLTK floating around, and it currently looks like another OS running inside an Intuition window. Something basic like that for X11 or GTK would be fine for now.

As regards shared libraries on RISC OS. Don't RM modules act as shared libraries? I always thought they did.

Chris

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@Chris_Y

Under RISC OS, filenames are very very different to what they are in Unix. There's an article I could refer you to, but the sordid details are not very interesting unless you use the platform. The point is that there's a comprehensive translation system in place between any Unix function that handles a filename and the underlying OS: fopen, stat, etc etc. This works in almost every circumstance, so could certainly be done for your library, and I don't believe Firefox makes any special demands of filenames.

fork() is a big problem too - the implementation we have is a long way from desirable. Still, most programs don't make real use of it, so we can get away with it. I don't thik Firefox uses it for anything important.

RISC OS relocatable modules are a long way from satisfactory as shared libraries. Their access priveleges mean they effectively run as part of the OS, and building an interface so they can act as a shared library is a great deal of work. It would be much much better to be able to directly compile libraries in the same way we currently do with static libraries and immediately have a shared library. Consider that we are dealing with at least 50 commonly used Unix libraries.

@Chocky

I'm quite familiar with RISC OS filenames, I used to use it a fair bit. In fact, if I didn't use Amiga, I'd probably own something running RISC OS.

All that would stop me buying one, if for some reason I do decide to leave the Amiga community*, is the little detail that I don't want to go from one "dead" platform to another. Especially when - and correct me if I'm wrong - the other platform is more expensive, less well supported and I don't have an extensive existing software base or the knowledge to make it worth my while. I do often cite RISC OS as the way things should be done (when they haven't already been done better in AmigaOS, that is ), so you've got me as a supporter of the platform in spirit, if nothing else.

Good luck with your project btw. The screenshots are looking good already.

Chris
* I'm not thinking of leaving

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

And it's out...

I wonder how long it'll be before we can see a similar article about AmigaOS? Looking at the Amizilla mailing list, things are definitely happening, although I won't spoil it for them

_________________