Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
The Next Big Step

By Armaphine in Technology
Thu Feb 15, 2001 at 06:26:27 PM EST
Tags: Software (all tags)
Software

It has been an oft-lamented fact that, for the most part, no operating system is ever perfect for both the sysadmins and the users. But by the same token, we must all have our own ideas as far as what we would like to see. What would be ideas that you would like to see in the Next Big Step?


An example: If your company is running Windows, then the users are happy (somewhat), since they have an OS that they are familiar with. They know, they have it at home. But you in the back office are running yourself ragged, trying to figure out how to plug all the security holes, and dealing with the at-random outages. The suits are not pleased that we have to keep buying 800 copies of a new OS every two or three years at $200-$500 a pop.

Alternately: If your company is running Linux, then the sysadmins (assuming they are proficiant in Linux) are happy. They have their ideal network, but the users are now griping. They are not leet. They do not know how to use Linux, and most likely have never even seen a Linux box before you set it on their desk. Whole new problem... your help desk is now swamped with calls asking everything under the sun. And the suits are also not happy, because the meeting that they had with BigCorp International fell through because MS Office couldn't open their StarOffice documents.

So what is the solution? Every OS out there has its pros and cons, but could it be possible to combine only the best parts of each, while still minimizing the cons? And what about other features that no OS right now offers, but would be of great convience? What follows are just some of the ideas that would need to be addressed...

DISCLAIMER: My specialty is with Windows 9X/NT. My memory of the MacOS is over five years old, and my experience with Linux is extremely limited. If I have made factual errors, I apologize...
  • Ease Of Use:
    If this mythical OS is to succeed, it must be easy to use, or it would get slaughtered. The main thing behind the entire Apple platform has always been that it was easy to use. It would let you do what you wanted, without the interface forcing you through its own hoops. Along the same lines, the Windows interface has succeeded the way it always has: through brute force. It has been out there for so long that it has burned itself into the engrams of everyone's mind. If you sit the majority of the computer-literate public down at a standard, default setup Windows box, within minutes, or even seconds, they will be playing solitaire.
  • Security:
    Obviously the need for security is there. Any little script kiddie with an internet connection, a handful of canned hacks, and a dab of youthful curiousity can start cracking into Windows boxes, whereas a well-secured, well-designed Linux system will take far more work, patience, and knowledge to root. And while you may have all the diligence required to protect your system, will your users? They'll set their password to '12345678' in a heartbeat, simply because it's easy for them to remember, and no amount of memos from the sysadmins will ever change that. While the password settings can be modified to make them enter line noise, you then run into a different problem: most people cannot remember a completely random password. You then have the problem of the users flooding the help desk with forgotten password trouble calls. A good middle ground must be laid out, between the need to keep a secure network, and the smooth operation of the buisness as a whole.
  • Process Control:
    If you work in industry, then you're probably very intimate with this subject. There are often these PCs that are connected, usually via a parallel or serial port to some sort of machine. The OS on them is probably some version of DOS, and the program controlling everything is probably completely undocumented. The question becomes: how could this new OS convince some of these companies to finally get rid of the older systems and move up to the new OS? They have no real motivation to upgrade... upgrading costs money, isn't sure, disrupts the schedule, and the old system works anyways. What can we offer these people to make them want to upgrade?
  • Business Applications:
    Let's face it. No one will touch an OS with a ten foot pole unless it's got a good office suite. VisiCalc made the PC a useful office tool, and WordPerfect put it on every desk, whilst Harvard Graphics made the suits happy in conferences. Nowadays, MS Office has a firm lock on the market, while StarOffice and Lotus Smartsuite gives it a run for its money. If this OS is to survive, a top-notch office suite needs to be included, bundled with the following:
    • Word processor
    • Spreadsheet
    • Database
    • Presentation creator
    • Web browser
    • Mail client
    And then, you have other applications that are a "must have" for any system to be remotely popular:
    • Anti-Virus (on Win32/MacOS at least)
    • Compression tool (i.e. WinZip)
    • Image / Movie Viewer
    • Games
    Why games? you might ask? Because if games are there... good games, something really good, will draw in a younger market. Younger market will probably lead to more home sales, since the parents will generally be trying to figure out the whole computer market through their children. Put all of these items together, possibly bundled with the OS, and the user base is greatly increased.
  • Internet Connectivity:
    Windows 9X made a big deal about getting connected to the internet, and with good cause. Internet usage has exploded over the past five years, from being amazed at seeing an http:// in small print at the bottom of a commerical, to the dot com boom. The question is: How will this OS connect to the internet? Obviously the question of the web browser is there... what to base it off of? Use Netscape as a model, Mozilla, Internet Explorer, or something else? Will web browsing and FTP searches be integrated into the basic look and feel of the desktop, or will it be in a class by itself? Will it incorporate instant messaging in itself, or leave it to third parties like AOL, Yahoo, and ICQ?
    And although security has been covered, security on the internet is almost in a field of its own. Will a person feel safe sending their credit card number, or social security number out through their computer using this OS? Or will greater steps have to be taken? What of encryption? The current standards are often broken far too easily, but then again, an apt hacker can pick apart the code and watch it all take place. From that point, all breaking the codes would be is a matter of reverse engineering. How would this OS handle that?
  • The Open Source Card:
    It's not exactly going to come as a surprise that many people would advocate this being an Open Source OS. But what about if the company making this wants to turn a Microsoftish profit? Free speech, free code, and free beer are all fine and good, but the suits will not be happy campers if we're just charging for tech support and packaging. So I humbly submit this idea: Remember Doom? The true A-List game of its time, and still a great play. Well, you could purchase the base game from id, and then download whatever freeware apps you wanted to make your own levels, sounds, graphics, and whatnot. You could practically turn the whole thing into a new game entirely, as long as you paid for the original copy. So this would be my idea: you purchase the OS for a price. In exchange for this, you gain the right to modify the OS to your hearts content. Want to build in better encryption? Go for it. Want to hack your web browser? Go for it. Want to change the second letter in fsck? Go for it. There would be only two stipulations:
    1. By doing this, you are hereby taking responsibility for the system. You will not recieve tech support from the OS's maker.
    2. If you wish to distribute your modified OS, for either fun or profit, you will do so in the form of a patch. The patch belongs to you. The OS belongs to its makers.
    In this way, the Open Source people would be happy, since they can once again have the right to make their own programs, and do what they will with their OS. It would also have the added bonus, I believe, of creating a secondary market for modifications to the OS. Anything from just modifying the look and feel, to simplifying or expanding the interface.
  • Separate or Together?:
    In my admittedly limited experience with Linux, it seems to be a very powerful OS, in the regards that it can run as a server, a workstation, or a stand-alone. Would this be the model for this perfect OS, or would it follow more along Microsoft's model, where the servers have one version of the OS, the workstations another, and the stand-alones and home users a third. Would there be issues in using the same OS for servers as home users? Or would there be a needless seperation between them, causing only conflicts?
  • Networking:
    How do we want our LAN/WAN set up? Do we want many small LANs, or do we want a few large LANs? Or alternately, could we combine the two into a multi-echelon setup? I know that NT is more apt to allow something along the lines of tying multiple LANs together through trust relationships. Also, Banyan VINES, if you are familar with it, was set up to have a few large LANs, then subdivided into smaller LANs, the resulting accounts coming out something like JOHN Q USER@TTP PROJECT@TOKYO OFFICE. The large LAN (TOKYO OFFICE), subdivided into smaller ones (TTP PROJECT), with members on each being able to log in from anywhere in the world, number of hops permitting. Would this be the ideal setup for a WAN, especially a global one, possibly with more subdividing? Or would a WAN acting almost like a LAN, with everyone being in the same domain be preferred?
  • Plug and Play:
    Better known as Plug and Pray, it has been held to a higher standard than it really was capable of, but the theory behind it was sound: Let the computer realize that you have new hardware installed, versus you telling it. But can it be improved upon? Or is it about as good as it's going to get? No matter what, it would be a much-desired addition to the ideal OS, especially for those that remember what loading device drivers was like before Plug and Play...
  • Roaming Profiles:
    On our perfect OS, we would probably wish to carry our personal settings wherever we went. Such things as mouse sensitivity, desktop wallpaper, sound effects, and such. Would these things be stored on a central server, or would a copy be downloaded to each machine you logged into? Also, might this OS make the settings available as a file, so that if someone had customized their settings in a certain way at home, that they might be able to carry them to work with them?
  • Portable Computing:
    The idea of picking up and going on the road presents new challenges for the perfect OS. From the networking aspect, how easily can it be for a laptop to pick up from one LAN and go to another? And how well will this OS run if it is loaded onto something like a PalmPilot or a palmtop computer? Will there be any changes that need to be made, or would an entirely new OS be needed, a la Windows CE?
  • Other features:
    • We have generic CD-ROM drivers and generic video drivers. The rest of the drivers, especially if in a corporate enviroment, are usually on the network. So is it not time that a generic network card driver was made? Something that doesn't have to have an enormous amount of bandwidth, just enough compatibility to load the necessary network protocols, log in, and snag the proper drivers.
    • Multiple desktops, a default selection on Linux, if I am not mistaken, would be an awesome thing to have built into this dream OS. For those that have worked on NT networks with some of the more strict security settings, often one must have the user quit everything that they are doing, log out, have the sysadmin log in, fix their problem, and then log back out, and allow the user to log back in to see if the problem has been fixed. The ability to just switch screens, log in with an administrator account (root or otherwise), change the necessary settings, then pop back over to the user's screen would be a great timesaver.
    • The ability to lock one's desktop with a simple keystroke would help improve security, even if it was just a small "coffee break" type password. Coupled with the idea of multiple desktops, it would allow someone to take a quick coffee break while someone else hopped on momentarily to check their e-mail.
What ideas might you, the members of kuro5hin, have? After all, the design of our operating systems affects all of us.
Had to get here somehow, you know...

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Related Links
o Kuro5hin
o Also by Armaphine


Display: Sort:
The Next Big Step | 33 comments (28 topical, 5 editorial, 0 hidden)
Portable Computing (3.50 / 2) (#2)
by nospoon on Thu Feb 15, 2001 at 12:40:10 PM EST

I just wanted to make a comment on Portable Computing. I just had Windows 2000 Professional Installed on my laptop and I have to admit that is handles going from one network to another very well.

I 'hibernate', unplug from my docking station and go home.
Once I get home or on the road, I plug into the phoneline to dial in to work or dial into my ISP to browse the internet. At home I can also plug it into my home network and it automatically detects the change and sets the proxy to my Direcpc Connection on my main box.

I have not had to 'log off' and back on or actually 'shut down' my laptop in weeks. I hibernate and resume and only have to enter the BIOS password to get back into Windows in about 20 seconds.



'Desire that is Friday'
Virtual Machines (3.66 / 3) (#3)
by SIGFPE on Thu Feb 15, 2001 at 12:44:35 PM EST

I would dearly love to see something along the lines of VM on IBM mainframes. The key idea is that every OS runs on a virtual machine (something like VMWare) whether or not it's the only OS installed on the machine. VM allowed you to have multiple OSes running simultaneously on a single machine and multiple virtual machines with various OSes running within other virtual machines. I'd like to see PC hardware and drivers designed with this in mind from the very beginning so that the hit from virtualisation is minimal. That way we need to argue much less about what OS we use because we can run them all!

I'm not an expert on VM so I look forward to hearing corrections...
SIGFPE

Take a look at VMWare (3.00 / 2) (#11)
by fossilcode on Thu Feb 15, 2001 at 01:57:30 PM EST

Check it out at www.vmware.com.
--
"...half the world blows and half the world sucks." Uh, which half were you again?
[ Parent ]
I've played with vmware... (4.00 / 1) (#12)
by SIGFPE on Thu Feb 15, 2001 at 02:27:03 PM EST

I rate it as the most amazing piece of software I have seen in the last couple of years. But it's still pretty clunky and slow. I'd like to have the facilities that vmware requires built into all OSes at a low level so that I can hotkey between OSes without taking any performance hit. It could be done...but possibly not with existing OSes and hardware.
SIGFPE
[ Parent ]
This would be quite what you expect. (3.00 / 1) (#23)
by plirr on Thu Feb 15, 2001 at 09:14:06 PM EST

VMWare doesn't quite do what I think you think it does. It emulates the hardware so that the OS software doesn't know it's not the only being in the universe. What you are describing would be an OS of it's own. Further you are positing that someone make what are currently OSes something like a pre-application layer. This would be a bit redundant, since the OS's whole point of being is to interface with hardware.

[ Parent ]
LOL (3.00 / 1) (#28)
by retinaburn on Fri Feb 16, 2001 at 08:33:46 AM EST

Your from IBM Sales and just trying to offload a couple dozen mainframes arent ya ;)

I think that we are a young species that often fucks with things we don't know how to unfuck. -- Tycho


[ Parent ]
Actually... (none / 0) (#31)
by SIGFPE on Fri Feb 16, 2001 at 01:01:32 PM EST

...I learnt about VM while working for IBM! I was pretty impressed with the way people could play with multiple OSes on one machine.
SIGFPE
[ Parent ]
The "let's do everything" syndrome. :-) (3.80 / 5) (#5)
by regeya on Thu Feb 15, 2001 at 12:45:46 PM EST

I'm a bit of a Linux zealot, so I'm going to approach the problem from that angle, but the same can be said of other operating systems.

Look at different Linux distributions. Red Hat uses as it's main selling point that they have a tech support infastructure. Debian's main "selling point" is that it has apt-get. Slackware's main "selling point" is that it's a simple yet more-difficult-to administer system. And so on. Yet each distribution has one thing in common: they try to cater to everyone.

At one point, I had considered putting together my own distribution (I'm gonna gloss over the details here) that included KDE, X11, some of the Red Hat autodetection stuff for hardware...and as I wanted to set it up, have all the autodetection stuff do 99% of the hardware setup. There would be no Apache, no ftpd, no telnet (or ssh for that matter) not allow remote X11 connections...and have init fire up X11 under a shared userlevel account. Essentially, something both as secure and as insecure as Windows (well, the personal Windows OSen) or MacOS.

And heck, distributions could do this. Mandrakesoft, are you serious about the desktop market? Don't ship Apache or any of the rest of the large number of network services (well, keep Samba...;-) Red Hat, are you serious about the server market? Don't ship systems with KDE, GNOME, or XFree, for that matter. I'm serious. Dump the desktop stuff if you're serious about the server market.

Anyway, enough of a rant for now, but I hope you get the picture. :-)

[ yokelpunk | kuro5hin diary ]

I think I do... (3.00 / 1) (#6)
by Armaphine on Thu Feb 15, 2001 at 12:50:08 PM EST

Again, being a 9X zealot more than anything, I didn't even know that the different distrubutions did all the different things. But the thing I was looking at was that the NT platform has a definitive seperation between Server and Workstation. I had thought that Linux was kind of homogenous between them, but it would appear that there must be a seperate OS for server and workstation...

Question authority. Don't ask why, just do it.
[ Parent ]

Different NT Platforms (none / 0) (#30)
by MrAcheson on Fri Feb 16, 2001 at 12:16:37 PM EST

The only real difference between NT Server and NT Workstation is licensing. Don't kid yourself that one is optimised above and beyond the other for specific jobs. The real issue between the two is how much money MS is making off its clients.


These opinions do not represent those of the US Army, DoD, or US Government.


[ Parent ]
The right tools for the right job (4.80 / 5) (#9)
by BehTong on Thu Feb 15, 2001 at 01:16:19 PM EST

OK, I'm sorry if I sound like I'm flogging a very dead and worn-out horse, but I think the problem you observe arises from trying to have one OS that fits all. As has been oft repeated and recited and chanted and clue-by-four'd, there is no such thing as one-size-fits-all, only one-size-fits-few.

What's wrong with using both Windows and Linux in the corporate setting that you describe? The two can communicate using SAMBA or what-have-you. Run your database servers and web server off Linux or another stable OS; let the desktop machines use Windows. Then the sysadmins can just firewall the Windows machines to keep them away from problems, and be contented with the better security of the server machines. Meanwhile, the non-tech people can stick with the familiar Windows, and not have to moan and groan and foam at the mouth for having to put up with vi or emacs or insert-your-hated-CLI-app-here. And they can use MS Word to their heart's content, and not have to worry about compatibility problems with the company's Most Important Corporate Client.

I personally do not believe there ever will be an OS that can meet everybody's needs and still be secure and powerful enough. You can only have the cake and eat it too up to a certain point. Cakes larger than that cannot be had and eaten too. :-) (I.e., after a certain point of complexity, everything has a tradeoff; you can't have everything you want.)

Beh Tong Kah Beh Si!

Why (3.00 / 1) (#10)
by rednecktek on Thu Feb 15, 2001 at 01:39:34 PM EST

Power, flexibility, security

Pick any two.

Just remember, if the world didn't suck, we'd all fall off.

Well, Linux is There for me. (3.50 / 4) (#13)
by Inoshiro on Thu Feb 15, 2001 at 02:28:51 PM EST

You mention having standards. That's great -- buy a bunch of EEPro100s from a local OEM dealer and get a discount. Those cards are excellent quality, one of the few good things Intel does.

You mention user settings going everywhere. Well, that's what a home directory is all about. If I blow away my drive and reinstall Slackware, my home dir survives. I don't have to customize a single thing. I may have to reinstall a few system-wide apps, but all my save games, theme info, writing, code, etc, are all there. NFS mount it, or use CODA. Add a nice GDM login on all workstations, and NIS exporting of user accounts, and you have the network where your settings do indeed follow you.

A lot of people in the Windows and Mac world hunger for these features. They come up with them, and proclaim that they are good. Then they pine away for them, knowing that in the great blob that in the Win32 registry, separation of SYSTEM and USER data is almost impossible.. All they have to do is migrate to a UNIX. Then they get it all.

As for usability, some of us open source programmers are working on it. We hear the cries of the oppressed ;)



--
[ イノシロ ]
Actually... (4.00 / 1) (#15)
by Armaphine on Thu Feb 15, 2001 at 03:22:38 PM EST

What I was referring to was, like under Win9X, you use OAKCDROM.SYS, and it'll load up any IDE CD drive. Windows also installs the VGA drivers automatically. It doesn't know what's there, it just uses something low-quality enough to adhere to pretty much everything.

All I want is a single, silver-bullet style NIC driver that'll just get it on the network. After that, everything else is usually saved on one of the servers, or can be pulled off the internet.

Question authority. Don't ask why, just do it.
[ Parent ]

Not gonna happen (4.33 / 3) (#17)
by fluffy grue on Thu Feb 15, 2001 at 04:04:59 PM EST

The reason that VGA and CD-ROM drivers can be made generic is because most video cards are VGA-compatible, and almost all CD-ROM drives conform to the ATAPI specification; the only reason that CD-ROM drivers are ever 'different' is because the driver specifically locks out drives which don't respond with one particular vendor name in the initial query.

Network cards, on the other hand, have very little standardization, outside of standardized chipsets such as the NE2000. NE2000 cards almost always ship with a generic driver, and Win9x comes with a generic driver, and so forth, which will work with any NE2000-based network card, for the same reason that any SCSI drive will work with a generic SCSI driver and any ATAPI drive will work with a generic ATAPI driver - the implementations are standardized and the interface must be adhered to or else it's not SCSI or ATAPI or whatnot.

This is also why you don't see millions and millions of device drivers for hard drives, and why under Linux there's just a lot of generic drivers for things like 'ne2000 PCI' and 'ATAPI CD-ROM' and 'IDE floppy' and 'SCSI CD-ROM' and such - because even though the manufacturers are all different and they like their DOS/Windows drivers to only work on their specific drive, the actual drives are all compatible; the Windows driver specifically locks out a competitor's hardware.

USB devices are all the same way, as well - they have to conform to a standard software-level USB interface (unless they are something which is specific-purpose) or else it won't be blessed as being USB. This is why most USB mice, printers, floppy drives, hubs, network devices, and so forth only require a single driver (even under Windows), and it's only things which aren't standardized under the USB spec (such as RS-232 convertors and digital cameras and such - and in the case of digital cameras, a lot of them do the Right Thing and just make themselves look like a generic storage device) which have driver divergence.
--
"Is not a quine" is not a quine.
I have a master's degree in science!

[ Hug Your Trikuare ]
[ Parent ]

Actually not (5.00 / 1) (#19)
by Skippy on Thu Feb 15, 2001 at 04:17:07 PM EST

That oakcdrom.sys does NOT work with all IDE CD-ROM drives and if you happen to have one that it doesn't work with you are COMPLETELY SCREWED. (Unless you are technically savvy enough to look up the Model number, visit Panasonic's site, track down the driver AND have another computer you can hack an alternate boot disk with.) I speak from experience. I'm not at home so I can't tell you the model number of the Panasonic drive but it works just fine under FreeBSD and Linux and won't give oakcdrom.sys the time of day.

Some of the newer graphical linux installs do basically the same thing you are talking about. Looked at a Mandrake 7 install? 640x480 VGA install with pretty GTK graphics. (That's not an endorsement BTW, just a statement. I think FreeBSD's installer is about the best out there and that is an endorsement.) I don't know about ethernet cards but I don't think windows has a lowest common denominator ethernet card driver either.

# I am now finished talking out my ass about things that I am not qualified to discuss. #
[ Parent ]

Not to criticize... (3.00 / 1) (#21)
by Armaphine on Thu Feb 15, 2001 at 07:37:55 PM EST

But I was speaking theoretically.

I realize that OAKCDROM.SYS won't pick up every CD-ROM drive, but it will pick up almost every drive. Sounds like you just got stuck with a really proprietary one. And as far as the VGA drivers, yes, they do load up pretty much automatically across the board.

As far as the NIC driver goes, no, there is no "one size fits all" driver out there for them. That's something I would like to see implemented in this fictional "dream OS"

I apologize for not making that point clearer.

Question authority. Don't ask why, just do it.
[ Parent ]

It's not about a driver, it's about a standard (4.00 / 1) (#22)
by plirr on Thu Feb 15, 2001 at 09:07:25 PM EST

The thing that you are not seeing, my friend, is that NIC's are not standardized. A single driver can not accomadate the array of NIC's out there. There is currently probing software that does just what need though, it probes the the system to check what NIC you have, and applies the appropriate driver. This is the case with Linux and Windows at least. I assume there are similar programs in other Unices.

[ Parent ]
+1 FP with one nitpick (3.50 / 2) (#16)
by wonderslug on Thu Feb 15, 2001 at 03:54:56 PM EST

My nitpick here has to do with your "Networking" section. Your LAN and WAN terms seem somehow to be related to NT's idea of "domains". I wish you had explained this more explicitly, because I had to read through the section several times before I figured it out. LANs and WANs are usually defined in terms that relate to layer 2 and layer 3 devices (within the OSI networking model). This has always been one of my pet peeves about NT - setting up a workstation to log into a server that is not on your LAN is not intuitive at all. It can be done, but I've only bothered to do it once or twice because it's such a pain. Networking is a topic that covers a vast array of technologies and you only touched on one aspect - basically just workgroup organization. As someone else stated, I think it would be wonderful to discuss all of these subjects in more depth. There's no way to tackle them all at once in a forum like this.

Change the country in my email to its initials if you want messages to get to me.
Here's what needs to happen: (2.50 / 6) (#18)
by rebelcool on Thu Feb 15, 2001 at 04:08:12 PM EST

Operating System specialization. Face it, today no operating system is ideal for all instances. Linux is great for network stuff.. I hate using it for anything else though. When I want to surf the net or get my more home-based generic computing things done, Windows 2000 it is.

Linux should focus on the server side of things. Windows should focus on business groupware, as well as a home market, and MacOS should continue its focus on just-learning-the-computer market.

Just as the programming world is beginning to specialize today, so should the OS world. It would make for better software all around (the linux programmers could fine-tune their network code...windows programmers on their general interest applications and so on).

In these respects, no OS is truly dominant over the others - EXCEPT in certain fields. Those fields have already become clearly defined. So now lets focus on them.

Of course nobody listens to me, but just wait a few years and see.

COG. Build your own community. Free, easy, powerful. Demo site

good idea, +1 (3.50 / 2) (#20)
by nickp on Thu Feb 15, 2001 at 04:20:44 PM EST

I think this is a very useful article as food for thought, regardless of whether we agree or disagree with ideas you presented. It's absurd to start nitpicking about what "operating system" means. It's clear that you're talking about a working environment. Microsoft Windows is marketed as an operating system, but any normal user would tell you you're not just buying VM algorithms, TCP/IP protocols, etc., but a whole environment. I am currently working with some people on some of the same ideas that were presented in the article in hopes of beating the traditional tradeoff between stability and usability.

I disagree with the application of the "right tool for the right task" argument over preference of work environments. An environment is not a tool. It should provide suitable atmosphere for a variety of tools you could choose for a variety of tasks.

"Gravitation cannot be held responsible for people falling in love." -- Albert Einstein

Why we need open standards (4.25 / 4) (#24)
by bjrubble on Thu Feb 15, 2001 at 09:54:37 PM EST

I see quite a few people saying that a single OS appropriate for everything is a pipe dream, and I think this is spot on. I'd like to see an examination of the assumption that we need it all in a single OS.

The one great thing that Microsoft and Intel did for computing was break the vertically integrated business models that IBM and others had imposed for decades. Decoupling the software from the hardware meant that the dominant player on either side had little power on the other -- you could buy hardware from IBM and software from DEC, or whatever combination worked best for you. And when that IBM hardware died, you could buy something from HP and just keep on going.

But now Microsoft, Sun, and others are leading the world into a vertical software model, where you buy your OS, database, applications, security, and network all from the same company. You can't choose the best of each category, you can't mix and match systems within an organization, and you can't swap out vendors later on.

The only places this strategy doesn't seem to get a foothold are where there are open standards. I have no hope of reading Word documents under Solaris, but as long as there's POP3 I can read mail. That's what I'd like to see in my organization, an infrastructure built on open standards that was accessible to whichever particular system best met each individual's needs.

Heh...I think we all think like you at some point (4.33 / 6) (#25)
by extrasolar on Thu Feb 15, 2001 at 11:14:52 PM EST

I think we all think like you at some point.

I think it comes to reading way to many advocacy groups. You read a cumulative amout of around a thousand posts regarding a certain OS, how it is good and how it is bad. Then you mind starts organizing these things...because thats what minds tend to do. And you start thinking, "I like Linux because it is more stable and has better security," when, in fact, you've rarely had any problems with your current OS crashing and your security problems are of the realm, "How did that company get my email address?"

So after reading so many posts, you end up recieving, subconsciously, an experience that you cannot account for. You read thousands of gripes about security, user friendliness, and then it almost seems like *you* are griping the same way, and are having the same problems they are.

So then what happens, is that you see pretty much every OS in existance as sucky in some manner. Even though you personally never experienced most of the suckiness. In addition, you begin to design your own OS...at least mentally. Because of this cummulative experience with all these posters on the internet, you benefit from the opinions and have a knowledge on how to design an OS that less people would gripe about. So you start saying about how this OS will have good security, without sacrificing performance; will be very user friendly, yet not sacrifice the kind of programmable power that power users expect.

But then...you run into trouble.

For every attribute your utopian OS has, there are 'n' different ways of implementing it. For performace, you can code the entire thing in machine code or low-level C...or perhaps just the low level routines. Of course, the high level routines will be coded in a high level language like Lisp. Of course, you may not know anything about programming anyway, but that doesn't really matter in your own mental OS. Because where you are getting your opinions of your design is from that cumulative experience that you can not account for.

But you can never judge a technology beneficial to use enough without sacrificing on some other attribute. Like power, stability, security, user friendlyness, etc. etc.

So finally, when you go through all these steps which may take months or even years, you turn full circle. You realize, for what it does, Windows 2000 isn't such a bad OS after all.

Either all the above, or I am the only person who ever experienced this. Yes, I know its lame. But its true. I have learned a lot of technology through the process. But sometimes when I post, if I don't catch myself, I may start speaking of things I don't really know anything about. But from many of the posts I read in this and other forums, I think I am not alone.

One size fits all (3.00 / 1) (#26)
by tfrayner on Fri Feb 16, 2001 at 02:14:02 AM EST

Just taking the argument to its logical extreme...

After reading the comments here, it occurs to me that the question of whether it is possible to have one OS do everything you might ever want basically assumes at least some kind of open source model. Otherwise you have a proprietary system (controlled most likely by a commercial entity) which could conceivably be responsible for the running of every computer on the planet. At least hypothetically.

Now I'm going to have nightmares :-(

Open Source vs Proprietary (4.00 / 2) (#27)
by retinaburn on Fri Feb 16, 2001 at 08:29:58 AM EST

I think it could work with a Proprietary OS if the company was 'resposible' to the user base. A company that would say 'ok we fucked up heres a FIX PACK, were sorry', or on the phone 'Well shit, look at that...hmmm i better get that fixed for you, thanks'. The code will still be locked up but they would understand that the customers are important.

Perhaps a co-op type idea where you buy the product you are a share-holder (per se). You get a voice because you own a piece of the company.

Course REAL shareholders would be pissed ;)

I think that we are a young species that often fucks with things we don't know how to unfuck. -- Tycho


[ Parent ]
Resource Scheduling (3.00 / 1) (#29)
by Bad Harmony on Fri Feb 16, 2001 at 09:16:13 AM EST

My pet peeve with existing PC operating systems is the way the CPU and network bandwidth are scheduled. Even though the speed of CPUs and networks have improved by orders of magnitude, one foreground task can make the other tasks almost unusable.

A proper CPU scheduler would guarantee each task a specific number of CPU cycles every millisecond. This would allow all tasks to be highly responsive to the user and external events.

The networking stack also needs to be able to offer bandwidth and latency guarantees to multiple tasks. With current operating systems, one FTP transfer is enough to make all other network dependent tasks almost unusable.

5440' or Fight!

point by point (5.00 / 1) (#32)
by bruce on Fri Feb 16, 2001 at 08:55:36 PM EST

I feel compelled to comment on this point by point, but first, let me say that Linux has made it farther than I ever expected. While Linux was good, I never thought it would make significant inroads against Microsoft. Now it's pushing Windows out of the server market, and its desktop UI and user install process are gaining rapidly on MS in ease of use -- so much so that, despite the conventional wisdom , I expect Linux will start displacing Windows from the desktop market in a few years.

You may laugh, but back before the net made it big, I didn't expect TCP/IP to be the one to span the world, either. I expected Compuserve or IBM or some standards consortium to come up with some other system, and TCP/IP would be displaced as a pioneering but obsolete academic/research protocol. Shows how much I know...

Anyway, I agree that there's an enormous amount of improvement possible in various aspects of OS design, both in the basic design an in tools and user interface. Eventually, everybody wants to design their own OS, and while Linux sucks less, eventually we're going to do better rebuilding from the ground up, and we might as well start now. So, point by point:

  • Ease Of Use
    Please keep in mind that the UI is (or, in a well-designed system, should be) largely independent of the kernel and other infrastructure. Once a GUI shell is developed for one OS, it can easily be duplicated on others, as Apple, Microsoft, and Unix/X have amply demonstrated.

    Anyway, I think we should take a cue from designers of Palm, Tivo and other embedded UI systems. While we don't need to throw away the WIMP interface altogether, there's no reason why a user should need anything more complicated than a 4-button cursor pad to control the GUI shell and system utilities.

    The control interface for the main UI shell and system tools needs to be uniformly presented. Windows has 4 or 5 different ways to view the filesystem, and none of them behave the same. It has even more different ways to view non-file hierarchies; the similarities in the objects need to be representable by the OS, and transparently viewable by the UI tools.

  • Security
    Security needs to be built from the ground up. That's one reason why Windows security sucks so much -- security wasn't a consideration when they started, so it had to be retrofitted at every stage of the game. Unix is better, but to be really secure, all applications need to run in OS-enforced sandboxes that give them no more permissions than are absolutely necessary.

    Package installation also needs to keep each instance of an installation in its own sandbox with a well-defined and well-controlled interface. Rather than scattering files, executables, and database entries far and wide with root/administrator privilege, the installation needs to be removable by simply deleting its directory.

  • Applications, Networking/etc
    By definition, applications aren't part of the OS proper. Since starting fresh is the whole point of the exercise, we're counting on blatant superiority to inspire application ports. The best we can do in this respect is to make porting easy.

    However, the notorious way to break into the business is the famed Killer App. Without speculating on which applications will be needed years down the road, if it's going to be an effective booster for a new OS, the App needs to use the OS's advantages in ways other OS's can't exploit. I'm thinking a cleanly designed distributed database will figure prominently in this respect.

    Microsoft knows this; as near as I can tell, their ".Net" and "C#" initiatives are a plan to tie everything into a networked database of their own design. However, because neither their OS or their database are very clean, I expect this experiment to fail miserably. If it doesn't collapse from its own weight, it will certainly fail on security grounds. You'd think they'd know by now that the net is a hostile environment...

Finally, I think Open Source is an winning attribute in a widely-used operating system. For sheer economic efficiency, they can't be beat -- and frankly, nobody deserves to make Microsoftish profits on anything. In the long run, Open Source will dominate any market where everybody needs a copy, particularly those with network effects. Proprietary OS's will never disappear, but they will be relegated to niche markets, where Open Source's advantage is not so great.

Mac OS X (none / 0) (#33)
by scruffyMark on Sun Feb 18, 2001 at 03:16:40 PM EST

I know I am going to sound like a Mac zealot, ranting away like they always do, and I suppose that is true to some extent. Nevertheless, everything the article mentions, and a good deal of what subsequent postings call for, sound just like Mac OS X (from my experience with the public beta, and a minimum of corporate propaganda). To mention the most salient points:
  • Ease of Use - well, it is a Mac, isn't it?
  • Security - well, it is BSD, isn't it? The hacked together OS underpinnings of earlier Mac OS's are gone. Darwin is a mighty nice UNIX OS in its own right.
  • Business Applications - MS Office will be available pretty soon (if it's not already). Aside from that, there is a pretty good selection of alternative office programs for the (classic) Mac OS, most will likely be ported sooner or later. Sun claims they will have StarOffice available this summer, but I'm not holding my breath... Compression is of course present - it just took someone writing a nice front-end to tar, gzip, zip, etc.
  • Open Source - well, sort of. Darwin - mach, BSD, the IOKit driver architecture, NetInfo as a management tool - is all open source. The higher layers - the PDF graphics engine, essentially most post-NeXTStep stuff, is closed.
It has some other nice features that I've been playing around with, but I've declaimed from the Mac pulpit enough for one day.

The Next Big Step | 33 comments (28 topical, 5 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!