[Deej]

Who would be interested in 802.11g wireless drivers? Interested in them enough to front some cash? And who would be willing to develop them? Are there any 802.11g PCMCIA cards out there with the needed specs available (or source)?

I have a small bonus coming this month, and I'd be willing to pitch in some cash to buy a developer an 802.11g PCMCIA card and access point, if others were willing to pitch in as well. I won't pay for the developer's broadband access though.

Anyone else interested in participating in a fund for 802.11g?

Deej

[h_ank]

I would, but I don't know when I'll have cash as I'm still unemployed. Also, DaaT, I'll give you a buck if you get rid of that emoticon menu which takes up half of my screen.
I'm thinking US$50 for the wireless networking driver could come from me.

[haiqu]

Deej,

A few thoughts on this.

1. Suppporting 802.11 or 802.11b would be cheaper for the initial hardware. The core of any drivers written would still apply to 802.11g

2. It will almost certainly never work properly without Bone. I spent months investigating this about two years ago. So really, until either Zeta or OBOS are released, forget it mate.

I have a stack of wireless hardware here already which I can't use, and was in negotiations with Be Inc for some development help when the sh*t h*t the fan. The targeted application was a Wireless ISP package based on BeOS. I guess this information doesn't need to be a secret any more, since it won't happen now.

Rob

[haiqu]

This is not to say that OBOS will include Bone, but it will have networking in the kernel, which kills the performance of the net_server.

Rob

[Deej]

[haiqu]

>> Not by much these days - I checked recently, which got me to thinking... the Linksys 802.11b/g (both) wireless cable/dsl router is in best buy for $139USD. A 802.11g PCMCIA NIC can be had on eBay for around $50-60USD == $200USD.

Linksys routers need Linksys cards, they don't support any others.

>> You can't get an 802.11b router AND NIC for less than $100, so, it's not all that much more. Besides, there are certain Prism chipset 802.11b cards out there that will work, I can't find one for sale at the moment, but I'm still searching...

I just picked up a Raylink 802.11 Access Point with NIC for US$90 a few weeks ago. 5.4GHz stuff is very directional, and even the fabled 802.11b "10MHz bandwidth" drops to 2MHz at about 20 feet so there's little difference for real world applications.

OTOH if you just need to cover one room, go Bluetooth.

>> I hate to say it, but I think just about everyone still around in the community is using BONE by now, no longer just those of us that had it legally provided.

Sure. But anything written to use it has no commercial future, which is why I dropped the project; believe me, it's a LOT of work. The "official" BeOS has no support for USB under AMD mobo's, no PCMCIA card services, and dead awful networking to name a few of the more _obvious_ limitations. Damn them for not officially releasing 5.1 to the world before expiring.

>> Aside from that, using the OpenBeOS networking component inside R5 might do the trick, if the developer of the driver was able to.

The developer of the driver would also need the new kernel to be working first, which is what he's been concentrating on since late 2001.


>> I'll end up with an 802.11b card to use with DSL, but looking long term, we should try to stay on the new hardware spec power curve, ya think?

No. Bzzt, wrong. You have to crawl before you can walk, and the Linux code for older cards is fairly stable, which is a good reference point. Getting hardware information for 802.11g would be a nightmare. I couldn't get support from Raylink even on their old 802.11 and I'm their distributor for Australia, so there's hardly any point even writing these guys.

>> I'm not saying we'd have a driver in a couple of months, but if we pitched in money now, got the gear for a dev, we'd likely have it when the OBOS net stuff is complete and fully ready to drop into a PE/Pro build.

Money won't make it happen any faster. IMNSHO anyone who has the skills to do this ought to get their priorities straight and pitch in to making the basics of OBOS work. We can ice the cake after we've baked it.

Rob

[Deej]

[haiqu]

>> >>Linksys routers need Linksys cards, they don't support any others.

>> This I did not know. 802.11b ones too? That wouldn't make much sense, I thought Linksys brags about 802.11b standard? I know that the box for the 802.11g said "draft standards", so that one I can believe.

The trouble with wireless is that everyone uses proprietary interfaces in the hardware world. 802.11(b/g/whatever) only describes the wireless link. It ain't as easy as it looks, but then nothing ever is.

>> >>The "official" BeOS has [...] no PCMCIA card services

>> This is dead wrong. I have 2 PCMCIA cards in my laptop.

Read what I actually said - no PCMCIA card services. Supporting a couple of common cards does not equal a PC-Card API for us to use as programmers.

>>But, the point of all this - I want wireless PCMCIA and I want it now, damnit! I just found a PCMCIA modem that works with BeOS, bought it, and 3 months later, after 2 1/2 years of waiting lists, DSL is finally available in my area - back to square one.

Yes, I hear ya buddy. I invested thousands of bucks into hardware I now can't use. Bummer.

>>[edit to explain my motives a slight bit more ] I can do dialup from any room in my house - there's phone jacks galore in there, but I'll be tied to Cat5 in one room with a DSL modem/router, and since my kids are home more than me and my wife could care less about internet, it'll be in one of their rooms with their computer.



There is a workable wireless solution for 5.0.3

See the data in this link, which I wrote a few months ago:

http://www.linux-hacker.net/cgi-bin/UltraBoard/UltraBoard.pl?Action=ShowPost&Board=bnews&Post=740&Idle=0&Sort=0&Order=Descend&Page=0&Session=

Rob

[haiqu]

Deej,

A bit more general data for you about wireless "standards".

1. Before 802.11 there was chaos on the 914 meg band.

2. 802.11 at 2.4GHz defines two speeds, 2Mbps and a fallback 1Mbps. On top of that it defines two protocols, FHSS (frequency hopping spread spectrum) and DSSS (direct sequence spread spectrum). It also defined two modes, ad-hoc and infrastructure. This therefore gives us a combination of eight variations assuming the rest is kept identical, which it isn't.

Vendor implementations of these protocols were not compatible with each other. For instance, my Raylink 2Mbps 802.11 cards are only compatible with Buslink (a Raylink rebadger), Webgear Aviator 2.4 (out of business now), and Breezecom. But Breezecom runs at a non-standard 3Mbps and only works with Raylink at the 1Mbps rate.

3. When 802.11b came along, some of the manufacturers saw the error of their ways and formed the so-called "wi-fi" consortium. This allowed a whole bunch of cards to interact at the wireless level at least, but only supported DSSS. This is unfortunate, because it hogs available channels and is very susceptible to noise interference.

802.11b supports 10Mbps with fallbacks at 5Mbps, 2Mbps and 1Mbps. They mostly don't work much faster than 802.11 unless the nodes are in the same room.

On top of this, some cards have no security, some support 40-bit encryption, some support 128-bit. In other words, the "feature" race made them all incompatible again.

4. Along comes 5.4GHz. Expect more chaos. It works at 20Mbps but reports of a 50Mbps "standard" are rife. On top of this, the laws of physics tell us that for an increase in frequency, two things will happen:

a) The signal becomes more directional
b) The range drops dramatically

Unless you plan to support an infrastructure network with RF amplifiers and towers, you'd be better off with a 10/100 NIC. Really.

On top of all the above, no version of Windows that I've tested supports routing an ad-hoc network through a second (hardwired) NIC into a DSL link, cable modem, satellite card or anything else. So you _have_ to set up an infrastructured network, which gets mighty expensive. That's why I wanted to do the WISP package in the BeOS.

Rob

[haiqu]

>>But once there, what happens? You are back to where Be was pre R4. Very little support, very little user base because of that. Keeping current with some commonly used, popular hardware, even if only one or two of each type, is very important, or you'll have an OS that will run on the OpenBeOS team's own computers, and that's all. I'm sure you get my point here.

Once we get a working kernel, libraries and basic hardware level drivers, they can be dropped into a standard 5.0.3 build. The fact that OBOS is binary compatible ensures that. So we lose no functionality and don't go back to pre-R4 at all. We only have to rewrite Be's drivers if they're faulty or lacking in features. Along with that, we can then write drivers knowing the underpinnings of the O/S, and thus do a proper job of it.

Supporting future hardware is simply a matter of writing code which complies to the current R5 API (as defined in the gnupro-991026 headers) in places where the current API is adequate. Where it isn't (such as multi-audio, USB, PC-Card services and so on) having the kernel sources available is vital to doing the job at all.

Rob