I saw on another thread (about networking) the phrase ‘everything on bebits will probably be packaged with Haiku’ in reference to drivers. As I understand, this means that a user will have about a hundred drivers installed that they don’t need. I have to say this disagrees with me. I can see the advantages, obviously, the it-just-works angle and all, but what about disc space, memory use, et cetera? Is there something in the Haiku kernel that makes this less of issue re: memory or is it just a ‘Moore’s low will fix it’?
I saw on another thread (about networking) the phrase ‘everything on bebits will probably be packaged with Haiku’ in reference to drivers. As I understand, this means that a user will have about a hundred drivers installed that they don’t need. I have to say this disagrees with me. I can see the advantages, obviously, the it-just-works angle and all, but what about disc space, memory use, et cetera? Is there something in the Haiku kernel that makes this less of issue re: memory or is it just a ‘Moore’s low will fix it’?
Drivers are modular and only loaded as-needed when the correct hardware is present. Gobloads of drivers are in BeOS R5 and it doesn’t have any particular issues with memory usage as a result.
Additionally, BeOS drivers are traditionally tiny. No 25MB sound driver packages with their own effects, control panels, status bar icons, etc, etc; more like a 25KB driver.
With small driver which are loaded only as needed, installing everything available seems prudent (for "just works"). But, if nessecary the option could be added to isolate larger drivers or localizations for install "Install **** Printer Drivers" or "Install Japanese Localization Support".
I saw on another thread (about networking) the phrase ‘everything on bebits will probably be packaged with Haiku’ in reference to drivers. As I understand, this means that a user will have about a hundred drivers installed that they don’t need. I have to say this disagrees with me. I can see the advantages, obviously, the it-just-works angle and all, but what about disc space, memory use, et cetera? Is there something in the Haiku kernel that makes this less of issue re: memory or is it just a ‘Moore’s low will fix it’?
Waldemar was referring to network drivers specifically - and it makes sense to include as many network drivers with an OS as you possibly can. Afterall, it’s via a network card that MOST people get access to the internet at this point, and once access to the internet exists, a user can download or update drivers from a remote resource.
Somehow I doubt there are “hundreds” of network drivers on bebits - and I bet if you started looking at the size of them, you’d have no issue with this.
Update: 33 network drivers listed on bebits - many of them are just patches (which don’t count as separate drivers)
And actually, if somebody thinks they have a bunch of worthless drivers, like printer drivers for really old dotmatrix printers, they can just delete it, unlike Windows and Linux. But, as MYOB said, that will only save disk space, memory usage will be unaffected.
Perhaps boottimes will be modified, since BeOS/Haiku has to load all of the existing drivers to see if they match any hardware? Will this ever become a problem? Would it be possible for the driver-loading part of the kernel to set attributes for each driver describing what pci ids it matches, and at boottime run a query for any drivers that match the ids for a device in question plus any unid’d drivers? Like, the 3c905 or whatever network card driver would have an attribute saying that it supports device x from vender y, and when a device gets found, the kernel would look for any drivers that have matching vendor ids and/or device ids and those that don’t have any ids? Would that help boottimes, because a driver would not have to be loaded unless it doesn’t have any ids in its attributes, and those ids would only be set once at the first time the driver is loaded?
But then, if all of the drivers on BeBits make BeOS boot in 30 seconds instead of 20, I guess it isn’t that much of a problem. Especially for R1.
–Walter Huf–
I think new devices (with the exception of low-level ones like disks or display devices) would be loaded into the hardware profile when the system is already up. Making the system check for all new hardware at boot time would be time consuming right? Besides, you SHOULD NOT have to restart your computer to install hardware that is not internal. It should just work.
Haiku R1 will have a new device driver model that only loads drivers matching a special PCI ID, for example. This will speed up booting time and driver recognition and even allow for selecting a particular driver for every device. Of course, only drivers that were written for that model will benefit (e.g.: the IDE drivers). Old drivers will continue to be loaded the old way. Still, that’s not a big problem (ATM).
fanton, did you ever read about how BeOS handles devices?
We are not Linux-ish. We don’t ever want to have those over-complicated solutions. Our OS should be fully understandable without studying it for years.
Haiku does not need a daemon. This is done in the kernel. Haiku’s new driver system simplifies the process and makes it reusable. A daemon is IMHO very inelegant because it is an application that is loaded after the kernel. Especially the Linux solution of using a thousand scripts is horrible! Haiku’s hot-plugging mechanism will basically be the same as its detection mechanism. If you really think about it, there should actually not be any difference between the two. That’s much more elegant and light-weight than Linux. It JustWorks™ without users being confronted with it (scripts or daemons that again must be run from a script).
Userland apps can use the node monitoring API to watch /dev for hot-plug events. BTW, the node monitor is normally used to watch directories and files for changes. Since /dev is a directory, too, we simply reuse an existing API (instead of introducing yet another one).
fanton, did you ever read about how BeOS handles devices?
We are not Linux-ish. We don’t ever want to have those over-complicated solutions. Our OS should be fully understandable without studying it for years.
From his other posts… I’m not sure he’s ever used BeOS. Or at least not for any decent period of time. He just seems to want a Linux-alike thats not Linux.
fanton, did you ever read about how BeOS handles devices?
We are not Linux-ish. We don’t ever want to have those over-complicated solutions. Our OS should be fully understandable without studying it for years.
From his other posts… I’m not sure he’s ever used BeOS. Or at least not for any decent period of time. He just seems to want a Linux-alike thats not Linux.
I did use Zeta for a month or so, but you are right. You see it as a bad thing. If I never used BEOS that doesnt mean I cannot talk about Haiku anymore since only people who used beos are allowed to. How would you ppl get new users then? Sucks forcing people to think BeOSish. Each person brings something unique to the table.
I said before that I like Haiku is just don’t like the BeoS ghost haunting you ppl.
I did use Zeta for a month or so, but you are right. You see it as a bad thing. If I never used BEOS that doesnt mean I cannot talk about Haiku anymore since only people who used beos are allowed to. How would you ppl get new users then? Sucks forcing people to think BeOSish. Each person brings something unique to the table.
I said before that I like Haiku is just don’t like the BeoS ghost haunting you ppl.
I love people making good and innovative suggestions. If you find something in Linux that is really good, please post it here. We could add it. But please don’t make suggestions that don’t fit our target audience. Linux is not a desktop OS (no matter what the OSNews posters say). It is plain too complicated in almost all areas. Don’t only look at Linux. Look at what else is out there. Look at what universities are researching. Look at what smart people are working on for the next era. But also look at very early research about interfaces. They had a lot of good ideas that were not feasible at that time.
fanton wrote:
And scripts deamons are good.
So? Please explain. I don’t like it when people just say “it’s better” without any explanations. Why do you like it?
Let me explain: primarily, computers are tools. They should make me more productive, so I get more useful things done more quickly. For you, computers seem to primarily be a hobby. That’s fine with me. But I don’t see anything useful in hacking on configuration files for hours. That’s lost time that I spend alone in a basement. In the mean time I could have done something useful and maybe thought about how to create a new UI that blasts off our asses.
I said before that I like Haiku is just don't like the BeoS ghost haunting you ppl.
The stated goal of this project is to recreate BeOS. If you don’t like BeOS, you don’t like Haiku. End of discussion.
youre only bringing back the dead…by R2 there will be little binary compatibility, and a lot of non-be programs…and haiku will be better than beos…why would you JUST recreate beos? that sucks…you already have beos, why have two?
plus be’s ideal was very egocentrical. a la microsoft or macos. but mac is becoming a pure software company, and microsoft is dying, being slowly killed by opensource.
and your reply makes no sense. I said I DO LIKE HAIKU. I dont care if you see Haiku as beos, I see it as something better than Beos, something newer, something that keeps evolving, that hasn’t been stale for last 5 years or so.
and lastly i hope my reply doesn’t offend you. but you are wrong. youre not only recraeting beos, you are improving on it.
I believe this has been the idea from the very start, and many of us agree that’s the best route to take. It’s been discussed, and discussed, and discussed.
Don’t let the plan fool you, though. Haiku R1, though pretty much a clone of BeOS R5, will be better than BeOS R5. First, Haiku is open source. Second, there are bugs in BeOS R5 that Haiku R1 won’t have. Third, it will have extended features not found in BeOS R5.
And don’t forget that Haiku is not the only BeOS-recreation project; if you don’t like the route Haiku is taking, go to one of those other projects.
