Slashdot Mirror
Dangers of Typecasting OSes
bjb writes "An interesting article has appeared on Byte's site on the dangers of typecasting operating systems. The article talks about specialization and purposes of Linux and BeOS. " Worth the reading, from the POV of dangers of stereotyping ideas/objects, in thsi case, putting BeOS throughly into the A/V corner, and Linux into the serve corner.
I know that when money is brought up people will respond with variants of "but there should be so much more to programming than money" or, alternatively, "look at Red Hat." But neither of those really points to an open-source BeOS being viable.
There is more to programming than money. But the fact is that all programs have development costs in time, and that time is always paid for: the payment is often simply hidden. If you are a high school student, your time is paid for by your parents; if you are a college student, your time is paid for by your tuition. If you are Richard Stallman, your time is paid for by MIT. This software may be free in the "free beer" sense and the "free speech" sense, but both those sense of "free" refer to the end user--not the developers. In a lot of the canonical cases of free software--written at or around college campuses--the development is (significantly) aided by the fact that the developers are supported by the institution.
If this wasn't the case--if the developers were full-time software developers not supported by an institution or by unrelated employment--how do they make money? Just by charging for manuals and support? This isn't a proven model, and the less ongoing support your software needs--support, I might add, of a sort which cannot be simply had for free on the internet--the less viable it is. And, unlike institutions, a for-profit company must eventually recoup its development costs. The FSF doesn't, because your tax dollars are at work keeping emacs going.
And, yes, let's look at Red Hat. They have given back to the open source community; I don't want to look into the Pandora's Box of how much they're riding on other people's coattails, but the fact of the matter is that they don't bear the costs of Linux development. They only bear the costs of developing their own contributions. How significant is this when it's compared to a company like Be? If we look at the S-1 statements, we can see that Red Hat's development costs for all of 1997 and 1998 were less than Be's costs for the first three months of 1999. This is not because Be is inefficient; it's because they're doing nearly everything themselves.
So the relevant question isn't whether or not BeOS would get drivers faster if it were open source (obviously, it would), nor whether an open BeOS could survive the company and mutate in different directions (obviously, it could). The question is: how would an open BeOS recoup development costs for Be? Whether or not it might be more philosophically attractive to have the operating system open (something some BeOS engineers have said themselves), Be, Inc. is not MIT. There is a point in the future at which they must not only break even but turn at least a modest profit in order to stay in business.
"But an open BeOS is independent of Be, that's the point!" Yeah, that's a good business case for not opening it, isn't it? Some of the argument against BeOS I've seen here is genuinely philosophical (while I don't agree with the "I won't use anything but open software" viewpoint, I respect it), but a lot has to do with the $70 asking price. If it came with the source code for $70, would that change that attitude? I'm skeptical. When Red Hat 6.0 came out, a lot of the messages I saw about it were pointing people to the places selling CDs of it for $5. There's a lot of free speech talk, but there's a lot of free beer walk. It'll be an uphill battle to recoup the $50-odd million development costs of BeOS as it is; how can it be taken open source without making that battle harder, or even unwinnable? I don't really think it can... but I'd be interested in hearing people's (non-flaming) thoughts.
I have a story about BeOS
For a few months last year, I had been using BeOS as my primary OS on a PMac. It was clearly superior to Mac OS in almost every way. I had used Linux, but didn't particularly want to use it on my own machine.
One day I switched ISPs. My new provider used QMAIL as their MTA. Turns out the mail_daemon that comes with BeOS is buggy, and doesn't actually generate correct SMTP. Most MTAs can still parse it, but QMAIL is a little stricter. The result: I can't send mail from BeOS.
I reported the bug with Be's tracking system, but they assigned it a status of "Will Not Fix". This despite the fact that it's a pretty easy problem to fix.
So maybe they eventually got around to fixing it. I don't know, I stopped using it, having suddenly realized the importance of free software. That sort of bug just doesn't happen.
Christopher Thomas wrote:
And if I want to tweak kernels, I can still use Linux. BeOS doesn't have to be the *only* OS; however, without meaning to offend, I think that many Linux advocates _do_ want Linux to be the only OS. This is a Bad Thing IMO.
No Linux advocate that I have dealt with (me included) want Linux to be the only OS. What many of us want is for Linux to be able to do everything. There is a big difference, but because this would put it in competition with every OS, some people think we want them all to go away.
Competition in Linux is friendly. We implement features, you imlement features. We get users, you get users. If we get lots of users, that's great. If some platform only has five Linux users, that's fine too, particularly since I'm sure those five REALLY REALLY want Linux.
----
Open mind, insert foot.
I'm going to try to keep my head here but I just spent all day trying to write programs for Win32 and MacOS so I may come off a little harsher than intended... :-)
:-)
:-)
i.There's a lot of baggage on the developer who wants to develop cool apps
for the linux desktop, too. Even with GTK or QT, you're still writing to a
wrapper to a library to a library to a driver, just to get a reasonably usable
API.
I think that you're missing the point somewhere here, which is that this is a Good Thing. Let me repeat that: *THIS IS A GOOD THING*. [ in fact, this is how Win32 works as well..although much less elegantly.. ]
Here is the setup:
X -- this is a display server. It has a single purpose for existance: accept requests to draw stuff and carry them out. The network-socket method of performing drawing operations allows a very flexible system. Often one sign (IMO) of a good design is that stuff that takes lots of hacking and kludging to get working in other systems simply arises as a natural offshoot of the way a given system works. For example, X's network transparency -- you can fake it in Win32 and MacOS with third-party software; X requires no modifications or special-casing.
Xlib: A simple (AFAIK) wrapper library which makes X protocol calls nicer by wrapping them in convenience functions. It has no other purpose.
Window Manager: One of the most villified aspects of the whole system because there are so many of them. A window manager (follow closely here) manages windows. It draws cute little frames around them, lets them be moved and resized, etc. This is generally its entire function. This may be accomplished with direct Xlib calls or not depending on the preferences of the window manager's author. Note that these are generally hairy beasts that people looking for an easy API probably want to avoid anyway
Widget Set (Qt, GTK+, etc): These use Xlib to draw useful things on the screen, such as buttons, dialog boxes, etc. Most programmers will use these, and they generally (except for monstrosities like XForms) have a nice simple API. The underlying layers can be ignored for most purposes. Issues like the fact that GDK is just an Xlib wrapper aren't necessary for new programmers to immediately grok as long as they can understand how to use GDK. [ the reason for wrapping Xlib again is to provide greater platform independence ]
So, in sum:
X server - displays stuff
Xlib - lets programs talk to X server
Window Manager - manages windows
Widget set - draws cute buttons, scrollbars, etc.
The thing that marks this as UNIX is not any feature of the complexity of the interfaces -- they vary by interface -- nor is it a silly amount of layering -- Win32 does this..it has a window manager (hidden from the programmer) and the common controls are essentially a widget set which uses Win32 drawing calls (Xlib) to do stuff -- but this:
** EACH FUNDAMENTAL SET OF OPERATIONS IS CONCEPTUALLY AND PROGRAMMATICALLY SEPARATED FROM (although not independent of) THE OTHERS. ***
If you have the kind of godawful mess that the Win32 API makes, where all levels of functionality are mixed together in the same place, it ain't UNIX. If you have programs that try to do 500 different things poorly (think IDEs or mail readers that know about SMTP and POP) you're using programs which are not UNIXey. If the you're outside the aegis of UNIX.
Why is this good? Programs that do a single thing well are simpler, easier to write and debug, and easier to use in unanticipated circumstances.
I'm personally interested in the Hurd; it looks like they may be going to out-UNIX UNIX and probably eliminate the worst cruft at the same time. Remember what I said about features arising naturally from a design? In the Hurd, several features that people have been trying to hack into Linux 'just happen', the most notable examples being userfs-type stuff and [I believe] per-process mounts. This is a Good Sign, although the Hurd is currently about five or ten years from being ready for general consumption
Daniel
Hurry up and jump on the individualist bandwagon!
The following is an excerpt from a mail I sent to the author after reading his article:
Something that I think that this typecasting touches on, is the conception of the "user". What we may have is not the typecasting of OSs, but the typecasting of "users" in a way which does even more damage. If OSs are marginalized, maybe a company or two goes under, but if people are marginalized, then voices go unheard and people are no longer able to excercise power thru computers, wether it is making web pages, sending love letters, or writing Free Software. I find the alienation of people a much bigger threat, and something which not only the marketers, but the pundits for each OS or "user" propogate.
In a Windows only world, the "user" is a consumer, passive and accepting of choices presented to them by the market. They are not supposed to create, contribute or otherwise influence the development of the computer, and their use of it, other than thru the very hollow power of the dollar. In the Linux only world, the technocrat wields alot of power, the user is expected to share the mania of computers that the developers have. This is just as disempowering for many, because instead of being confronted with opaque surfaces, they are inundated with complexities.
The same thing happens in a BeOS only world, as it also has a very specific conception of "user". Something very similiar to Apple or Windows with a more sophisticated technical edge. We should be careful not to typecast BeOS as the third part of a Hegelian dialectic of the "user", the synthesis of two opposing concepts. We should not do that with ANY OS for that matter, because that action STILL will produce a single conception of the user which will alienate and isolate many people.
As Linus and others have pointed out, Linux world domination should not be about being the only OS, but about making the OS a choice, by allowing all different types of people choose the computing platform they want, and which suits their needs. These platforms should then be able to interoperate with one another, so that no one "user" ever becomes dominant. Obivously, this is not what the present personal computing market looks like, but I think it's something that we can all achieve.
craig
He's right, of course, about the mantra of Open Source being a powerful meme. Look at comment #1 in this thread. Linux is attracting a great many users because of the press it gets, but this user base will shrink when the honeymoon period goes away. As sorry as WinWhatever is, it is still easier to use than Linux, and it will continue to be this way for the next few years, IMO.
This does not diminish what Linux has to offer. People who love Linux will continue to love it, even after a lot of new users have returned to their original OS.
What the writer of this article neglects to mention is the vast amount of hardware that Linux will run under vs. BeOS, which runs on older Macs and a pretty good array of PCs. (Someone from the BSD camp should post a reply to this to point out that BSD is the most ported OS in the cosmos.)
Other reasons for Linux being more popular (right now): 1. Cheaper. You can try it out, then go buy a full distro later, or not at all. 2. More press. Microsoft hasn't started an anti-Be task force. No one has heard of Be in the mass media since Apple threatened to buy them.
Linux share will grow on both the server and desktop side, but until installation, maintenance, and UI are improved, most users will run after a few short days left alone in the Colloseum. Penguins bite.
While Slashdot reported that quickly, they didn't report the followup--that Blender is going to be coming to BeOS and that it was described by the porter as an "error in communication" which Be took responsibility for. The new GL library, to my knowledge, required the rest of Genki (the next OS release), which is still in beta.
As stated in my previous post, you don't need to recompile the kernel to add drivers to Be, or need access to the kernel source to write drivers. Most drivers are kernel modules loaded at boot-time, analogous to "modules" under Linux.
I have already said that I recognize that sometimes things _do_ need to be tweaked in the kernel - but I think that for the vast majority of _users_ who don't do kernel tweaking, BeOS is fine. Linux isn't going away any time soon, don't worry.
This guy starts out with a wonderful hypothesis: You shouldn't typecast OSes, you should look at their real capabilities first. Then he abandons that theory and typecasts Linux as a server-only OS.
He criticizes Neal Stephenson for portraying Linux and Be as a tank and a Batmobile, respectively, claiming that Be isn't a Batmobile because you can drive it to the store every day, but, oddly, has no objection to that characterization of Linux.
Well, guess what. I drive my tank to the store every day. I use Linux for all of my ordinary tasks, on both server and workstation. I stopped multi-booting ages ago, because Linux is just a flat-out better desktop OS than Windows.
Is it better than Be? I dunno. I looked into getting a Batmobile, because they do look very cool, but I didn't in the end. I checked the HCL on the Be website and discovered that in order to run Be on any of my machines, I would have had to replace, at the very least, motherboard, CPU, video card, sound card, and mouse.
So 99% (to use the number he apparently pulled out of his butt) of Linux users don't use the source. Well, that means the other 1% do. 1% of a few million is a fscking lot. And, thanks to free source, all those people can (and do) write drivers. That's why Be has (woefully short) lists of the hardware it works with, and Linux has (amazingly short) lists of the hardware it *doesn't* work with...
Be has my vote for becoming the OS/2 of the next decade. Technically elegant, theoretically excellent, and practically doomed by a lack of third-party support.
No Linux advocate that I have dealt with (me included) want Linux to be the only OS. What many of us want is for Linux to be able to do everything. There is a big difference, but because this would put it in competition with every OS, some people think we want them all to go away.
Um, take a look at some of the other posts in this thread. And in the threads attached to the MacOS X post. Many of the posters here, who would certainly call themselves Linux advocates (whether or not they fit the definition that you use), have a strong knee-jerk reaction that says that any non-Open operating system is intrinsically evil and should be abolished. They then proceed to loudly make this opinion known.
They also generally don't mention *BSD. When it comes up, the reaction is along the lines of "Huh? Oh, well, since it's Open, it's ok.". And then they go back to bashing non-Open OSs.
This seems to indicate that many of these "advocates" _do_ want to actively destroy non-Open OSs regardless of whatever merits they may have. And for many of them, Linux == Open == Linux.
I'm not saying that _all_ Linux advocates are like this by any means - but a substantial and very vocal minority are.
First, I have lurked around for a while, and I have been fairly impressed with linux, the users, and what can be done in the OSS model.
I think, though, that a lot of linux users are missing the point. Linux is a GREAT server OS. If I ever had to run a high traffic server I would run linux. But Linux is NOT a GREAT desktop OS. It can be a good one, but can anyone really say its great? (I shall define great as being able to be used by most users immediately after install). I have tried linux. I don't have the time to invest into setting it up or tweaking it so that it works. I have also tried Be. I was able to use it, go online, and start d/l software immediately without the help of a friend directing me. (btw I consider myself pretty technically literate, I am an MIS major with a CS minor). Be doesn't as yet have many apps for the general user, but that's because it is set up for being a Audio/Visual niche operating system. (Be is being smart. Remember they are a company that has to make money to put out a good product. Going head to head with the Redmond giant is suicide. Linux could do it because they don't need to come in on budget or have to make money) Look at linux 1-2 years ago. Where was StarOffice then?
A lot of people have stated that Be has no apps or they can't use the apps they have on linux. Just because Be is closed-source doesn't mean that you can't create apps for it or port apps to it. Be comes with a fully integrated BASH command line interface. It comes with a variety of GNU tools, most notably gcc. If you do some research you may notice that it is nearly fully POSIX compliant (they are not fully POSIX yet because they want to put their time into getting more hardware supported).
Have any of you checked out the specs on Be? Be uses a 64-bit Journaling File System. It has pervasive multi-threading, pre-emptive multi-tasking, and excellent use of SMP. It already has a security/multi-user architecture set-up waiting for a time when it will be usefull.
Many have said that they could not use X. Well if you go to the BeWine page you will see that they have ported X-Windows over and are working on a port of Wine as well.
I will finish with this last thought. Please remember that the OSS dev model transends all OS's. Just because an OS isn't open source does not mean you can't design a good app for it. The only reason there aren't a lot of OS apps on Be is because THEY HAVEN'T BEEN WRITTEN YET (and/or ported). Before you start thinking of how to flame me, remember that linux too was once really young with no apps, and that it took time before mainstream users started to notice it.
Let the moderation begin.
TC
Very well said (and perfectly correct). Unfortunately a majority of the people on slashdot don't seem to see it this way, and jist of their comments are rather offensive (or distasteful) to non-Linux users. (ex: your priorities are different from mine, therefore you are evil, stupid, lazy, crazy, and etc).
Sigh. Why can't we all just get along and be open minded?
The writer suggests that 99% of users couldn't care less about having access to the source code [of an OS]. He has it almost backwards here. If we accept his number of 99% and apply the remaining 1% to the 10 million or so Linux users, we are left with 100,000 people who *do* care about digging in source code. Thank you, Scot, for pointing out the mindboggling advantage that Linux holds over closed OS's.
I have no particular interest in kernel hacking. Ordinarily, I "couldn't care less" about digging in the source. However, when I found myself needing a tiny tweak to my server OS, I was *SO* glad I could just go in and change a single line of code and solve my problem.
Geeky modern art T-shirts
This article misses one of the most important features of Linux: you can use the software without signing any licensing agreement, or even paying for it. There's been a lot said about the virtues of "free speech" software, and I agree that it is beneficial. However, we shouldn't ignore the fact that "free beer" software is great, and that it's one of the main reasons that so much "free speech" software is available. Idealism is a fine thing, but it's hard to compete with free toys.
That's "Mr. Soulless Automaton" to you, Bub.
Boy is that true. He can't even play _himself_ without getting made fun of. And that's just about the only role he can get, too.
I'd like to see Adam West in some other stuff.
-- This and all my posts are in the public domain. I am a lawyer. I am not your lawyer, and this is not legal advice.
Just putting a bit of the blame at the feet of those who deserve it...
The reason I bring up OS/2 is because I think BeOS has a chance of avoiding what brought down OS/2. Granted, some of what brought OS/2 down is that IBM did a sloppy job of marketing it (or so I've heard). But another problem was that Microsoft successfully altered people's perceptions to see Windows as a "standard" part of their PC, and OS/2 a useful "extra", or an "alternative" OS. These days, Linux (with the help of Linux users and advocates) has been altering people's perceptions and they are realizing that Windows doesn't necessarily have to be their OS by default. It's my hope that the success of Linux will be the wedge to drive open the OS market, making products like BeOS available for those who just want their computer to work and don't want to spend fifteen hours poking around inside the nuts and bolts.
What I'd like to see is for Linux to bring about a new OS market in which competition can flourish. Of course, what I'd really like to see is for an open-source "category killer" to arise in the OS arena (similar to sendmail, for example) but that doesn't seem likely to happen anytime soon. In the meantime, I hope BeOS succeeds.
-----
The real meaning of the GNU GPL:
"The Source will be with you... Always."
How many Linux users take advantage of access to source code?
The OS source is like a "home defense" handgun--not essential but Really Nice To Have in certain situations. I find that having gcc around is the best thing about Linux. How many other OSes have free compilers practically built in?
BeOS, on the other hand, is immediately satisfying.
Immediate gratification may be the American Way, but generally, the more powerful things are, the longer they take to learn. Tricycles are easy to ride; bicycles are not. Guess which goes faster. Everyone here knows where Linux stands in this setting...
Give a monkey a brain and he'll swear he's the center of the universe.
We all hear things like "My pet rat who doesn't know the first thing about computers hates Windoze but loves the linux installation I gave it", but realistically, we're nowhere near the ease of use and fun for the average user provided by Be, or (gasp!) Windows.
Gnome is cool, KDE is nice, but maybe pushing linux into that space is the wrong way to go. Linux is wonderfully configurable, but it's still a unix at heart, and unix has way too much baggage for the novice to slog through.
There's a lot of baggage on the developer who wants to develop cool apps for the linux desktop, too. Even with GTK or QT, you're still writing to a wrapper to a library to a library to a driver, just to get a reasonably usable API. The sound system(s) are OK, but they certainly could be better. Contrast this with programming for BE, and it becomes pretty obvious that we don't hold up well.
What we in the open source community need is really a fresh start on the desktop. Linux is evolving wonderfully as a server and workstation OS for the power user, but if open source is going to win the average user, I really do believe we're going to need something very different from UNIX.
For now, though, I'm sticking with Linux since, though I think BE is a cool company, corporate control of an OS has repeatedly proved disastrous in terms of benefits for the end user.
Hacker isn't just some random writer for byte. He wrote The Beos Bible which has a pretty in depth coverage of the BeOS including descriptions of the internal workings. I think it is naive to say that he is just repeating what he has read elsewhere, because he has had far more contact with the developers of Be than most users have with their OS, open source or not.
Yes, it does. I have a BeOS machine sitting next to me right now.
While the OS is closed, you still get all of the header files and semi-complete documentation (what isn't there is being written). Applications can be closed or open as the applications programmer wishes; AFAICT, Be doesn't care. More applications are good regardless of the license they're released under, and Be is too small to write applications itself. It loses nothing from open apps, and loses nothing from making the entry points to the OS visible.
Java got a slow start due to it's 'web language' typecast, which gave M$ enough time to destroy what it could have been. Perl is strong enough for more than just scripting. C/C++ (with libraries) is full enough to cover scripting duties. Typecasting is just an easy for for PHBs not to have to learn or think.
int freemem = 1, *bill;
char *windows;
do {
windows = (char *)calloc( 1024, sizeof( char ) ) ;
if ( winmem == NULL ) {
freemem = 0;
} printf("haha %s", (char)bill);
}while (freemem == 1);
Only 'flamers' flame!
Um, no, by straight numbers most users seem to think that Windows is the best choice. You're trying to convince them that Linux is. He's trying to convince them that BeOS is.
How many computer users - out of the total number of computer users - even program? How many, given the opportunity and the tools, _would_ mess with the kernel? Very few.
Linux is nice. I just plain _like_ *nix variants. But the most that I've done under Linux is look in the
I've recompiled _applications_, but the kernel header files are all you need for that, and Be gives you those.
However, I'm getting sidetracked. While I enjoy Linux for programming, most users aren't programmers. In terms of installation and ease-of-use, IMO BeOS wins (though Linux is catching up).
Programmers would love Linux. Many of them already do. IMO, the users that *don't* program would love BeOS. They are who use most of the desktop machines.
Bear in mind that these are applications. There's nothing to prevent apps under BeOS from being Open or Free. It is only the kernel that's closed. Now, you could make the same arguments about the kernel as you do applications, but in practice I see fewer kernel bugs that need fixing.
And if I want to tweak kernels, I can still use Linux. BeOS doesn't have to be the *only* OS; however, without meaning to offend, I think that many Linux advocates _do_ want Linux to be the only OS. This is a Bad Thing IMO.
Re. drivers, I'm happily doing driver development under BeOS with no more access to the kernel source than any other user. The headers are all that's needed.
The people to take that up with are the BeOS users, not Be itself. Be makes an active effort to publish all of its API and document as much of it as possible. If you want to release an open application under BeOS, there's nothing stopping you. Porting isn't that hard either, because the guts of BeOS are Unixoid.
This points out a big problem with not having source code in terms of "typecasting users". If Be is the only company with the ability to improve BeOS itself, and since Be, as a small company, doesn't have the ability to improve it in every way, then they are forced to typecast their users into a few basic categories, simply to focus on what to improve next. These basic categories would hopefully encompass most BeOS users, but would certainly leave out large subcommunities.
Whereas if BeOS were open sourced, individual communities of users, no matter how far from the mainstream (such as the BeBox owners) would have the freedom to improve BeOS in ways that might only benefit them, and Be would never commit to doing themselves. You might even see other companies providing customized versions of BeOS, and a market of different BeOS flavors, like the different Linux distributions.
--
Jake
Why do people immediately have to scream Open up the BeOS source? It has many disadvantages as well as advantages :
1) It has controlled and organised design and strategy - Something the open source model doesnt do well is promote _designing_ systems and maintaining them. Can you see an open source project creating a GUI + API like Be's from scratch in anywhere near the time? They group would argue via email forever about details and eventually abandon the project in favour of another X port (because its avaliable)
2) Because it is commercial it can be adopted more widely than 'unsupported' (in the legal sense) free software by companies.
3) Hardware companies are happier to supply proprietry specs for writing support for commercial systems rather than free ones.
People should stop bickering about the 'principle' of it being closed source and take it for what it is and look at both sides of the story subjectively.
Peace,
~Pev