I'm a but frustrated that it seems the Prism2 driver is the only possibility to have wireless networking for AmigaOS. So...

I'd like to discuss, on as technical a level as we can, wireless networking and work on defining a good bounty to help bring support to AmigaOS. There's things I don't know much about, such as WPA/WPA2 details, what makes a good or deficient wireless API for driver writers, or driver/software differences between 802.11a/b/g/n. So I'd like to see if anyone here may know something to help make a better bounty definition.

I myself do not care if AROS or MOS are supported as well, as I have neither of them, but I won't argue against supporting them. Are there other efforts for those to combine with? Sharing scarce programming resources is a nice idea, but consider that if we don't want to differentiate ourselves, why are there three Amiga and -alike OSes instead of just one in the first place? I'm also concerned that simply porting Open1X (and other open-source items) is also part of lack of differentiation, and we might as well just use Linux...

I envision something that sits on top of Roadshow, handling wireless issues such as a GUI interface for selecting encryption settings, entering passcodes for protected access points, scanning for and selecting access points to connect to, choosing default connections when more than one are available, etc. as well as offering a common API to the wireless chip drivers, so the drivers can be as minimal as possible, not needing to reinvent things for each one, compared to the prism driver which I understand has all WEP built in to it instead of separating that for other drivers to benefit from.

This system at some level will need to interface with PCI and USB, are there other forms of wifi wireless "cards" to potentially use? Would this level of interface be better located in the common library, or would it be better in each individual driver?

The API should be well documented so anyone can write a new driver to work with it should their chip not already have one. The system should be well designed and extendable enough to well serve ANY 802.11 driver, and potentially other standards as well. No particular chip in a supported standard should require any special changes to this system to get it to work, that would indicate a deficiency of this wireless system.

I do think that anything currently possible for 802.11* on other platforms should be possible here. I think that it may be interesting to have some sort of feature plugin systemi, be it a direct plugin mechanism or via he use of other libraries. I think it may be suitable to consider encryption methods as plugins, one for WEP, one for WPA, as well as for other features such as one for Radius enterprise authentication, etc. Such plugin may possibly also support other things like WiMax, cellular network cards (with SIM cards or without SIM cards for us American CDMA users on Verizon/Sprint/etc) or external phone tethering, etc.

Does anyone have a good understanding of Radius? How important is it?

What about other 802.11 imporant keywords that I'm for the moment unaware of?

How do 802.11 A/B/G/N, WiMax, cellular, etc. differ from the driver and software perspective?

Does it make sense to try for a general wireless.library, supporting wifi, wimax, cellular, bluetooth, etc? Or should this forget about other things, and envision more of a wifi.library/802.11.library, then semarate wimax.library, cellnetlibrary, etc? each on top of Roadshow, sortof beside each other in that heirarchy? If split, what would still be common between them, such as GUI interfacing type stuff that could be shared, and begin with this bounty?

Someone mentioned Open1X, which is an open-source wireless system, dual-licensed as GPL or BSD. I'm not against porting existing software, but I also like the idea of differentiating AmigaOS from Linux. If all we do is port things over, then all we are is an awkward platform to run Linux software on, when it's easier to do that directly in Linux. But we do of course have a bit of a resource deficit compared to the land of Linux, so porting helps us get anything at all a lot of the time these days... But I myself would have preferred that to be LGPL. Open1X says it can work with any TCP/IP stack so long as it has access to a DHCP client, which I believe Roadshow should do quite well.

In terms of GPL vs BSD licensing of Open1X, what are the pros and cons of each in this situation? Does the GPL license forbid closed-source drivers? That's important, as some chips may come with NDAs to driver creators that forbid open-sourcing the driver. I'd prefer the ability to have as many drivers as possible, and would choose whichever license is not an obstacle to that.

If this ends up as something other than a port of Open1X (or any other similar open-source thing), I'm less concerned with how this system itself is licensed than I am with making sure anyone can write a new driver to work with it, and that as many chips as possible can have drivers written for them (considering possibility of chip documentation NDA forbidding open-sourcing its driver) . Though I myself would choose between closed-source or LGPL.

What chip vendors forbid open-sourcing drivers for their parts, if any?

OK, let the discussion begin!

Last edited by billt on 13-Jan-2010 at 01:23 PM.

_________________
All glory to the Hypnotoad!

@billt

Quote:

Does anyone have a good understanding of Radius? How important is it?

For an enterprise, Radius is very important from security point of view for AD environment.
Users log in from wireless AP side and gets verified against Radius before entering the intranet.

For home use, less so, Radius for home use is overkill.

Quote:

Or should this forget about other things, and envision more of a wifi.library/802.11.library, then semarate wimax.library, cellnetlibrary, etc?

Recall we discussed this some time ago.

Yes, for a start, a general wifi 802.11 library should be best.

As long as we don't have netbooks or laptops, what is needed is software able to connect to WEP 64/128 and WPA/WPA2 encrypted wireless network.

_________________
Site admins are people too..pooff!

@tomazkid

Quote:

As long as we don't have netbooks or laptops, what is needed is software able to connect to WEP 64/128 and WPA/WPA2 encrypted wireless network.

What is different about desktops vs laptops/netbooks?

I seem to have missed the late December thread. Off to read... My own previous one I only remember the result of my question was only Prism2 works.

Last edited by billt on 12-Jan-2010 at 09:05 PM.

_________________
All glory to the Hypnotoad!

@billt

Quote:

What is different about desktops vs laptops/netbooks?

Well, perhaps not much. But desktops most often can do with just 802.1x standards you create yourself, and since you make it yourself, you can make it compatible with your own equipment, and make the encryption for what you need.

A good support for all 802.11 networks and WEP 64/128 and WPA/WPA2 should be a good start.
When that works, other standards can come later.

_________________
Site admins are people too..pooff!

For the idea of having a common plugin API, however those plugins are implemented, how comparable are the data structures and values between WEP, WPA, and WPA2? Is a single API reasonable, or do different things need to be available for the different encryptions, Radius, etc?

I know now that Radius may not be as useful for most of us AmigaOS users, but I'd still like that to be considered during API design, so that it's possible later on without a huge rewrite of the initial wifi system. Similar for other features that I may not know to mention.

_________________
All glory to the Hypnotoad!

@tomazkid

Quote:

A good support for all 802.11 networks and WEP 64/128 and WPA/WPA2 should be a good start. When that works, other standards can come later.

So long as the initial wifi API is well designed so these other standards can come without much effort rewriting this system to fit them in. Should they not be named and considered during wifi API design to make them easier to fit in later? As well as for consideration in the GUI design for settings, data structures, save file, etc?

Last edited by billt on 12-Jan-2010 at 09:56 PM.

_________________
All glory to the Hypnotoad!

For those that would like a cross-platform wifi bounty, I'm ignorant of MOS and AROS things.

Do they both have compatible/similar SANA-II networking stacks? PCI and USB libraries? File access? Are these "native" APIs to each OS, or are they secondary APIs used only for "compatibility" to older software that hasn't been rewritten as native?

I've not done GUI stuff myself, are you happy with using the lowest-common-denominator in MUI compatibility across the three, or how hard would it be to have the GUI part in separate OS-native interfaces such as Reaction and whatever AROS uses?

What different relevant APIs are the same of similar?

What APIs are noticably different? What extent is the effort to overcome differences?

_________________
All glory to the Hypnotoad!

An old (1997) intorduction to wireless extensions in Linux
http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/Linux.Wireless.Extensions.html

More recent documentation and links to successor to LWE
http://linuxwireless.org/en/developers/Documentation/Wireless-Extensions

_________________
All glory to the Hypnotoad!

@billt

It's a nice idea (but unfortunately beyond my programming skills). However, at a packet level all the interface types should be pretty much the same, the only differences will be the initialisation and control of them. I guess a wireless daemon / preferences utility would be required, and this would present a single, standard SANA-II API to the system as well as a standard UI for detecting networks etc. That way there's a good chance it will work with most flavours of AmigaOS with very little modification. Then the individual plugins would handle the individual on-the-fly encryption / decryption, and also the hardware accesses as different chipsets seem to be accessed differently (otherwise the Prism driver would work for every card).

_________________
RobTheNerd.com | InstallerGen | SMBMounter | Atoms-X

What about devices like this:

http://www.asus.de/product.aspx?P_ID=9H8EmdWucHDpYOb8

It has various functions, most interesting is the function of an Ethernet adapter.

Quote:

Under WL-330gE´s Ethernet adapter mode, you can convert all the Ethernet-capable devices (printer, desktop,..) into wireless.

It's configurable with a Web interface and can be powered with a simple USB-cable, so don't need another power supply (and it's powersaving, too, since it only needs the 500mA supplied by USB). It brings my Mac mini with MorphOS to the internet via WLAN. Very recommendable.

@billt

We would need:

1) The roadshow interface TCP/IP "driver", ie the module that tells Roadshow what to do with the traffice

2) a generic Wireless interface, that allows you to set-up connections to wireless networks (with all the relevant security plugins)

3) Hardware drivers (ie drivers for specific chipsets - like the Prism one) to interface with number 2 above

Security:

The WPA/2 specification includes the Enterprise versions of each so while Im not a coder I wouldnt mind betting that to build WPA in you need to build the Enterprise part in (I have never actually seen any router/OS having one and not the other) as well. The Enterprise part is the Radius authentication that everyone has been talking about.

I would imagine that as usual number 3 would be the hardest part given the lack of documentation out there.

Mark

_________________
IceStar Media Ltd.

@_analogkid_

I also think that this sort of device is the easiest way to get wireless.

My brother had a ASUS WL-330 adaptor like that, but had some problems getting it to work.
So when I needed a wireless adapter as well we found the Linksys WBP54G.
It is quite cheap, but works great.
It's designed to be used together with the linksys IP-phone adaptor, but you can use it standalone as an accesspoint/bridge for any LAN based device.
You will need a 5VDC powersupply, as it is not in the package.

@_analogkid_

Quote:

It has various functions, most interesting is the function of an Ethernet adapter.

Standalone WAP-thingies are fine, and most often can do the job, at least as long as they can speak encrypted with each other.
(manufacturers use different algorithms every now and then, which gives separate results for the password, result = same encryptions are not compatible. Even Apple did this once, with one Airport upgrade, resulting in users loosing their wireless connections after the update. )
but, as long as you stick to one manufacturer, the above should not be a problem.


But, for those who want wireless built into the OS, it is not an option, and if we ever get laptops/netbooks, support for at least 802.1x is needed, together with the common encryptions.

_________________
Site admins are people too..pooff!

@_analogkid_

No, I want wireless support. If we ever get a laptop, I will not want to carry one of those things everywhere I go.

_________________
All glory to the Hypnotoad!

@amyren

Quote:

I also think that this sort of device is the easiest way to get wireless.

Yet it goes exactly nowhere in bringing wireless SUPPORT into the OS. Topic title updated to reflect this component.

Last edited by billt on 13-Jan-2010 at 01:24 PM.
Last edited by billt on 13-Jan-2010 at 01:24 PM.

_________________
All glory to the Hypnotoad!

@billt

If we ever get an official laptop, they're hardly going to release it without wireless support from the OS! If it were a possibility, I've no doubt that Hyperion will provide a driver and Wifi infrastructure to allow it.

_________________
RobTheNerd.com | InstallerGen | SMBMounter | Atoms-X

@tomazkid, billt, amyren:

It's clear that this kind of option is no one for laptops, and I'm pretty sure that there will be builtin-wireless support when the first portable device is supported by an Amiga-like OS (next one I see is Powerbook/iBook-support for MorphOS, but OS4/AROS can take some profit, too).
But for currently supported desktop-hardware, these adapters are quite a good solution, although the Mac mini has wireless support on board and makes this adapter "too much" for MacOSX or Linux.

@Daedalus

Quote:

If we ever get an official laptop, they're hardly going to release it without wireless support from the OS! If it were a possibility, I've no doubt that Hyperion will provide a driver and Wifi infrastructure to allow it.

Wouldn't it be great to have wireless support and drivers already there if someone finally comes along to make the laptop hardware?

_________________
All glory to the Hypnotoad!

@_analogkid_

Wile those external ethernet->wireless dongles may be one good solution for desktop users today, my intent with this topic and resulting bounty is to head toward OS level support and drivers for internal PCI cards and USB widgets, using the chips more directly, and with more variety of chips drivers possible and more encryption and speed standards than are available in the existing prism2 driver. Considering these external adaptors, and/or the existence of prism2 driver, as a complete wireless solution for AmigaOS is IMHO disappointing, and I wish to move beyond them. These external ethernet things are not beneficial to the intended goal here, however useful they may be otherwise, please continue discussions of those elsewhere.

_________________
All glory to the Hypnotoad!

Is it reasonable to ask that a wireless framework be capable of making the Amiga into a wireless access point (going to the internet via ehternet wire) for other devices to connect to over wifi? Or is that too much additional stuff to be part of an initial bounty?

_________________
All glory to the Hypnotoad!