the vast majority of people who actually have experience with it call it HP-SUX
I haven't used HP-UX for a while, but it was a very solid Unix version. I think the HP-SUX label is just from people who lack either skill, or imagination, or both.
It was noticeable that the GNU file utils were way ahead of HP-UX (and every other proprietary Unix), but the market of most interest was enterprise server, and the reliability of HP-UX in such a role was impressive.
The killer for me, was when HP started migrating its own internal systems to Microsoft Windows NT4. Where as SUN had Solaris everywhere in their UK offices, from the receptionists desktop, to their servers. I know SUN must have had some MS Windows boxes somewhere for testing, but you could wander office to office without seeing anything but Solaris.
Probably a misconception from a recent comparison that said by compiling more code into the applications, MacOS depends on less shared libraries, which it does. Of course that is why you need 15GB of disk space to install it, and twice as much RAM to make it run as quickly as GNU/Linux (and I assume regular BSDs would be faster than MacOSX in this regard as well), and the security updates are huge.
Everyone seems to have forgotten why we create libraries in the first place, and why we use shared libraries (I mean we do have a choice, well sort of, no one seems to use libraries that aren't shared out of choice these days).
You have web applications that rely on IE in order to work.
Well I don't, and the one fancy feature in our web server stuff at work died with Windows Vista, as they withdrew the control it relied on for "security reasons". Presumably these security reasons aren't bad enough to have switched it off in IE7 for XP SP 2.
Now the only thing I have at work that is browser specific is the firefox search plugins, I mean I could rewrite these using OpenSearch.
Relying on IE only features may result in it not working in the next version of IE (or Windows), which is why one uses open protocols, and programs in programming languages with competing implementations.
I think that's severely oversimplifying, because rewriting the system to take only one system call would certainly result in more bugs, no?
I thought Alan Cox had already done a kernel module for serving http?
But no rewriting the system to more specifically do the task in a more focused way would almost certainly result in a lot less bugs, of course the system would be less "generally useful".
Clearly it is a simple argument, less is more.
Backwards compatibility has huge costs, one of them is security. Supporting those apps with 8.3 filename limits, and 3 or 4 different ways of accessing the file system, all mean there is a lot more around to go wrong.
If you are actively using large chunks of "more" you probably don't care, as your system is more flexible, or more featured.
But I'm really not interested in the performance hits the more bizarre features of SMB gives to my webservers, but I daren't switch it off, as I know I'd be running an IIS configuration that is practically unique in the world, and it is flaky enough as it is. Similarly I don't care about that 8.3 compatibility, I know I could switch it off, but I'd worry something obscure might break. So I'm stuck with the "more" even when I want "less". Where as my Linux webservers don't have a GUI, most don't have SMB (or NFS), I lost all that network filesystem junk with the last update on most of them (scp (or http) will do fine for most things).
Guess it comes down to design - the secret of elegance is about what you take out, not what you put in.
And if you want (or are unsure if you need) binary backward compatibility to DOS 1 (or whatever level is provided), you can take out very little.
The Debian Iceweasel maintainer has previously noted that the Ubuntu Firefox was more different from Mozilla's than Debian's, and it is relatively easy to check.
Now it is entirely possible the Ubuntu folks have suddenly binned all their patches to Firefox, but I think you might be mistaking stuff read on/. for the truth;)
Of course it may well be that Iceweasel is more different now, but that probably reflects branding changes.
But I agree a few days here or there for some fixes (depending on their type) is likely neither here nor there. Heck IE manages to be vulnerable to known remote code exploits most of the time, and most people still use it in preference to struggling with installing software on Windows.
I thought the interesting thing was the idea of using Redhat as an Enterprise Linux desktop, I'd assumed that with Redhat 9 being dropped Redhat really weren't interested in the desktop market. Certainly it wouldn't be my choice for an Enterprise Linux desktop, as there are a number of derivative distros for which desktop is clearly their only or primary goal.
Microsoft have rather more than 700,000 getting updates automatically - although I'm not sure their model would be a good one to copy, when Google discover it doesn't work with a proxy server quite right, and you need a GUI on every box to run IE;)
Even the Debian repositories update more than 700,000 machines. But then they use a wrapper around rsync as well.....
Indeed the article suggests that the GNU/Linux package solutions like Redhat RPM, and Debian debs, may be overengineered for their needs. Since they don't need to support multiple architectures (yet!), i.e. there problem is less taxing than the more general one facing GNU distro vendors and other OS vendors.
However I think minus several zillion points for reinventing the wheel, given what a nice selection of wheels already existed for this particular problem, which would have allows reuse of third party tools if nothing else. Still a lot of them charge per CPU licencing (ouch!).
mapreduce would presumably make it very easy to count Google's servers in near realtime.
Finally a program using "mapreduce" that I could write without my brain hurting.
Re:Definitely has uses but..
on
Oracle Linux?
·
· Score: 4, Insightful
HP-UX use to ship with the kernel settings all correct for a small Oracle database (small then being 9 to 20GB IIRC). Of course being a conscientious system admin, it didn't stop me double checking them against the Oracle documentation each time in case some advice had changed.
This didn't happen by chance. But it meant that you could be reasonable certain no obscure kernel settings were incorrectly set (at least not by an oversight, didn't stop people setting the wrong settings when tuning).
At the time Oracle were talking with Hewlett Packard about a stripped down HP-UX to build "Oracle Servers" on PA-RISC. It made sense then, and it still makes sense, except HP-UX is no longer the "obvious choice" for an Oracle server.
To be honest, I think in the GNU/Linux world, it is choice of certified hardware that is probably as important, if not more so, for Oracle, than choice of distribution. Since I've been bitten by underdocumented, under tested, RAID hardware or Linux drivers for same (the effect is the same, no matter where the fault lies). If you are aiming for really high availability on an Oracle database, buying the solution as one stop from Oracle makes sense.
I doubt cost-wise it would be that competitive with DELL and Redhat, at least initially, but for some applications hardware cost is irrelevant compared to unplanned downtime.
Something like Debian, or Ubuntu, with long support periods, and completely freely redistributable base (with builtin rebranding -- "no Mozilla says you can't call it..." hazzles), is the obvious sort of base. Although presumably BSDs might be an option as well. Or Oracle might still want a big corporate backer for their distro variant.
> If I'm not mistake, you've spent a great deal of time with UserLinux and its predecessors in an attempt to solve the matter of a standardized platform.
There are reasons to standardise components that aren't related to package management. Such as end user, and system admin training.
My understanding is that this was where UserLinux was coming from. The people involved all seemed very happy with how Debian packaging worked, but concerned they couldn't possibly support customers using; "KDE" and "Gnome"; Perl and Python; Ruby on Rails, and Catalyst; Postfix, and Exim.
Of course I switch between Debian and Windows, between Perl and PHP and ASP, between sendmail, and Postfix all day long, and as a result I'm pretty useless at all of them, and half the time can't remember what the comment character for the language or format I'm working in now is (I must learn Emacs, it handles that for you).
UserLinux basically built a list of preferred applications (a meta package), and let the Debian packaging system pull in the stuff it needed to make the list work. Thus using the power of the packaging system, not trying to work around it.
The packaging complexity problem is not as bad as you make out, because a lot of activities by different packages are completely orthogonal. i.e. Debian developers know they won't clash on documentation file location if they get the package names unique, because it will go in/usr/share/doc/.
It is building this "orthogonal" nature into systems that makes them more robust and scalable. Which is why xinetd should be preferred over inetd, because you can configure it by adding a file, rather than fighting over entries in/etc/inetd.conf. So one uses dependencies when they save you redoing work, but design systems so that components don't interfere with each other unless the interaction is planned.
Surely the ultimate point is you can have any filesystem that fits in the Microsoft shaped filesystem hole, with free software you can change the hole. Okay 99% of the time it isn't wise to change the hole, but when it is Linus/Theo will recognise it and take the patch.
Anyway the other response on knowing "how it works", is summed up nicely if you go read the progress of some of the ReiserFS on Windows projects.
Hehe, surely the main difference with free software is you can run a similar quality tool over the source code, and know if the programmers involved sorted these kind of defects, or even just look it up at Coverity.
So you know if the program you are using is better than average or not.
I was actually pretty encouraged, I have Linux servers, running Apache, and Postfix, with Postgres backends, and Perl front ends, with a little PHP. Authentication via PAM.
When I looked at my deployed base of software, I'm seeing a lot of code with zero defects detected remaining (okay, a lot of them probably haven't been upgraded for reasons of system stability, if these fixes didn't make Debian stable, and I suspect a lot won't because they aren't that important).
Of course I know they are low defect products, that is why I chose them (without Coverity's help I might add). This leads the main security issues being my programming skills in creating web applications, and utilities. You don't need to use units as big as "per Kloc" when measuring my defect rate.
We also run some servers with Windows, IIS, various COM objects, and I can only assess the quality by how the software behaves, i.e. bugs affecting functionality, on that basis the equivalent proprietary software sucks in comparison, but is very clearly getting better recently. Of course these numbers say nothing directly about functionality issues, but I think experience tells us there is some correlation, otherwise how did I end up with so many low defect applications?
As others observed the published data speaks volumes more than the commentary. Since code can't have negative defects.
What no "qmail" statistics? Given Wietse noted some of the Postfix "defects" were false positives, I suspect it would cast more darkness than light, but hey they should test it just so we have something else to bitch about next week, now we've covered vi versus emacs.
Of course Emacs biggest defect is I don't have time to relearn how to use it, and that sort of defect wasn't covered either.
I switched a PC from Windows 95 to Debian, when Windows 95 went out of support.
Okay I'm not a typical PC user, but then I don't test any of the free software I hack on under Microsoft Windows anymore because I don't have a Microsoft Windows licence, and I don't see the point of paying for one (heck I just threw out the 120 day evaluation CDs for Windows 2000 Advanced Server I got at the launch party). So there were consequences for other Windows users of my switch, be it ever so small and insignificant.
So I don't think it is totally far fetched, but I doubt the Debian mirrors will crumble under the load.
Support is needed because there is software in Windows 98, that has known critical level security issues (I think unfixed vulnerabilities triggered by merely viewing images), so if they use the Internet, eventually they'll have to do something, if only reinstall and be more careful what images they view.
The great thing about kit like this is you can teach them how to dismantle and rebuild them safely without risking anything valuable. Guarantee you'll have the interest of most of them if you dismantle it, and explain the parts, especially if you make them help, or do it themselves. For kids dismantling things that are "expensive" is forbidden fruit.
PC technology has barely changed internally from a visual perspective, apart from SATA replacing the ISA bus, and the amount of fans and pins on the CPUs.
It won't teach them anything useful for when they leave school on specific hardware, but in the mean time they'll be able to repair their own (and parents) kit, learn how not to get electrocuted, how not to wipe out sensitive electronics with static, they can learn about why complex things are made from components, the importance of standards, what a transformer is, what fuses are for, what is in a computer. One can even bring in the topic of mass production, and why such complex devices can be produced so cheaply.
Who knows, if you have some decent kit, they might even learn what a Faraday cage is. Okay scrub that one;)
Finally they can learn about recycling, environmentally sensitive disposable, landfill, leaching of heavy metals, etc.
Do watch out for sharp edges -- but hey that is a lesson in itself, and they'll learn it quickly enough.
The folks who gave Congress power to lean on Big, Gigantic Corporations like Google are the people who fear Big, Gigantic Corporations. According to my scorecard, that's usually been a Left-Wing plank. The folks who try to limit Congress's ability to make life miserable for Big, Gigantic Corporations like Google are usually depicted as Right-Wingers.
I think the left/right thing will have to go. Even some very aggressively pro private enterprise, western governments have managed to end up with "bigger government" at the end of their term in Office.
But then I've yet to deprogram myself from the propaganda about free enterprise, and revisit the facts. There are a lot of unhelpful memes out there -- and a surprising number of them seem to be tenents of someones economic theory, and all of them are a lot less 'obviously true' than that axiom about parallel lines not meeting.
Governments should regulate markets, and prevent monopolies, if they do that well we've little to fear from big corporations. Only the extremists don't think governments should regulate markets, that is not to say some extremists haven't been influential.
"Harvard says that because MS has more market share, it will have more market share."
They should read Nietzsche;
"The wind in the valley and the opinions of the marketplace of today indicate nothing of that which is coming but only of that which has been."
Yes if business continues as normal, Microsoft will continue to have a dominant marketshare in the desktop operating system market. Anyone who thinks things will continue as normal hasn't learnt much from the history of the last hundred years.
The whole point is the technology is seen as disruptive, that means business isn't continuing as before. The current GNU/Linux desktop share is effectively as close to zero as makes no odds, but if you look at who are the several million Desktop GNU/Linux users you see it is a very odd bunch, including some big opinion formers. I'm not that optimistic for a GNU/Linux desktop -- in a sense it doesn't matter much to me, I have a perfectly good free software desktop operating system at home and work, if the market grows a little, especially server side, and so I get a bit better hardware support I'll be happier.
What will make GNU/Linux as a desktop, if it ever does, is countries adopting it, or schools adopting it. Probably the biggest issue is lack of GNU/Linux know-how. People just don't know how to deploy it, or how even to set up basic desktop behaviours. Sure a few companies and individuals know, but it is probably the leading limiting factor for both business and home use, even over software availability and lock-in.
The best place to put spam filtering is at the endpoint - that's where the most information is available to make the decision and the end user can intervene and provider feedback to the classifier (e.g. gmail). If I start filtering spam, in the hope of reducing the chances of being blacklisted, I will be doing a disservice for my users.
Absolutely disagree.
Best place to do spam filtering is at the termination point of the first SMTP connection. Okay in a perfect world this would be the end point, but the world isn't perfect.
User interaction is pretty much useless in spam filtering.
1) There is too much spam. 2) Users are too error prone (Trust me on this I have a SCOMP feed)
I have no choice I have to do spam filtering before the end point, because if I don't, the spam filtering at the end point will reject the email (and or blacklist my server), and it'll end up in my queue, and I'll have to try and return it (otherwise email would become unreliable), and some poor smuck gets 100,000 delivery reports for emails he never sent.
I have about 1000 domains on one of the email servers at work that forward the email elsewhere, and the average queue length is about 40 messages (with 5 day expiry). The queue consists almost entirely of spam from domains that won't accept email, refused by the final destination server. So I think I'm doing quite well on the spam filtering front.
We don't really do "false positives", I guess maybe a single report every few months, almost invariably due to misconfiguration of the sending email servers.
As no one with Debian uses it, or looks in it, but XaraLX is in "non-free" for Etch and Sid. Looks like the maintainer upgraded it to 0.7 in Sid, Etch is 0.6 currently.
The current schedule (AIUI) is for NetWare 6.5 to ship as a guest OS in Xen on SLES 10 mid-2007.
I assume the motivation is client driven. They may have ported a lot of their own code to GNU/Linux, but I expect there is a client base of installed Netware, with third party software etc. That said I haven't seen Netware running anywhere for years, but then I'm no longer consulting in a different place every week.
i presume the redhat dude meant was 'redhat isnt ready to commercially support xen'
The folks at Novell have more motivation.
They have para-virtualisation of this thing called "Netware" running under SuSE (hmm sure I have a dusty certificate somewhere saying I'm certified on Netware). It lets Netware run on boxes that Netware doesn't have drivers for. It lets customers consolidate servers, upgrade hardware, and keep running their investment in Netware, and I bet Netware is a lot simpler to get running reliably (well as reliably as Netware ever runs, I wonder if SFT works under para-virtualisation) under SuSE, than say a whole enterprise GNU/Linux distro.
Redhat spent "millions" testing Xen ?! Seems a bit much given how much goes into testing some kernel changes.
No one mentioned the deliberate mistake in the article, Ubuntu have to distribute the source for 3 years after they stop distributing the binaries, not "for as long as they distribute binaries".
I agree with the "it isn't difficult line". My Desktop PC has enough disk space to store all of the Debian mirrors and archives (and I'm not planning on doing a distribution with three kernels and 11 architectures), I might have to switch off mirroring, but the disks weren't the expensive bit of my PC. Since you can charge costs, all that matters is storage and organisation, storage is cheap, and storing the source code for stuff you distribute should be second nature for anyone with experience of source code management.
"DRM flight" (or perhaps we should label it the "freedom drain") is even better news, because that applies to Microsoft Vista as well.
Matt who did this Ubuntu on PPC thing earlier in the year http://mattl.co.uk/blog/2006/01/26/switcher-tales/ has just switched to Debian Etch, and says he thinks it is even better than Ubuntu. But hey they are both free, he has a choice, although it did take him some effort to get all his weird Apple formatted music to play. I suspect on that score he switched "just in time".
Slashdot Mirror
I haven't used HP-UX for a while, but it was a very solid Unix version. I think the HP-SUX label is just from people who lack either skill, or imagination, or both.
It was noticeable that the GNU file utils were way ahead of HP-UX (and every other proprietary Unix), but the market of most interest was enterprise server, and the reliability of HP-UX in such a role was impressive.
The killer for me, was when HP started migrating its own internal systems to Microsoft Windows NT4. Where as SUN had Solaris everywhere in their UK offices, from the receptionists desktop, to their servers. I know SUN must have had some MS Windows boxes somewhere for testing, but you could wander office to office without seeing anything but Solaris.
Probably a misconception from a recent comparison that said by compiling more code into the applications, MacOS depends on less shared libraries, which it does. Of course that is why you need 15GB of disk space to install it, and twice as much RAM to make it run as quickly as GNU/Linux (and I assume regular BSDs would be faster than MacOSX in this regard as well), and the security updates are huge.
Everyone seems to have forgotten why we create libraries in the first place, and why we use shared libraries (I mean we do have a choice, well sort of, no one seems to use libraries that aren't shared out of choice these days).
Well I don't, and the one fancy feature in our web server stuff at work died with Windows Vista, as they withdrew the control it relied on for "security reasons". Presumably these security reasons aren't bad enough to have switched it off in IE7 for XP SP 2.
Now the only thing I have at work that is browser specific is the firefox search plugins, I mean I could rewrite these using OpenSearch.
Relying on IE only features may result in it not working in the next version of IE (or Windows), which is why one uses open protocols, and programs in programming languages with competing implementations.
But no rewriting the system to more specifically do the task in a more focused way would almost certainly result in a lot less bugs, of course the system would be less "generally useful".
Clearly it is a simple argument, less is more.
Backwards compatibility has huge costs, one of them is security. Supporting those apps with 8.3 filename limits, and 3 or 4 different ways of accessing the file system, all mean there is a lot more around to go wrong.
If you are actively using large chunks of "more" you probably don't care, as your system is more flexible, or more featured.
But I'm really not interested in the performance hits the more bizarre features of SMB gives to my webservers, but I daren't switch it off, as I know I'd be running an IIS configuration that is practically unique in the world, and it is flaky enough as it is. Similarly I don't care about that 8.3 compatibility, I know I could switch it off, but I'd worry something obscure might break. So I'm stuck with the "more" even when I want "less". Where as my Linux webservers don't have a GUI, most don't have SMB (or NFS), I lost all that network filesystem junk with the last update on most of them (scp (or http) will do fine for most things).
Guess it comes down to design - the secret of elegance is about what you take out, not what you put in.
And if you want (or are unsure if you need) binary backward compatibility to DOS 1 (or whatever level is provided), you can take out very little.
The Debian Iceweasel maintainer has previously noted that the Ubuntu Firefox was more different from Mozilla's than Debian's, and it is relatively easy to check.
/. for the truth ;)
Now it is entirely possible the Ubuntu folks have suddenly binned all their patches to Firefox, but I think you might be mistaking stuff read on
Of course it may well be that Iceweasel is more different now, but that probably reflects branding changes.
But I agree a few days here or there for some fixes (depending on their type) is likely neither here nor there. Heck IE manages to be vulnerable to known remote code exploits most of the time, and most people still use it in preference to struggling with installing software on Windows.
I thought the interesting thing was the idea of using Redhat as an Enterprise Linux desktop, I'd assumed that with Redhat 9 being dropped Redhat really weren't interested in the desktop market. Certainly it wouldn't be my choice for an Enterprise Linux desktop, as there are a number of derivative distros for which desktop is clearly their only or primary goal.
That isn't what Debbie told me.
Depends how you look at the problem.
;)
Microsoft have rather more than 700,000 getting updates automatically - although I'm not sure their model would be a good one to copy, when Google discover it doesn't work with a proxy server quite right, and you need a GUI on every box to run IE
Even the Debian repositories update more than 700,000 machines. But then they use a wrapper around rsync as well.....
Indeed the article suggests that the GNU/Linux package solutions like Redhat RPM, and Debian debs, may be overengineered for their needs. Since they don't need to support multiple architectures (yet!), i.e. there problem is less taxing than the more general one facing GNU distro vendors and other OS vendors.
However I think minus several zillion points for reinventing the wheel, given what a nice selection of wheels already existed for this particular problem, which would have allows reuse of third party tools if nothing else. Still a lot of them charge per CPU licencing (ouch!).
mapreduce would presumably make it very easy to count Google's servers in near realtime.
Finally a program using "mapreduce" that I could write without my brain hurting.
HP-UX use to ship with the kernel settings all correct for a small Oracle database (small then being 9 to 20GB IIRC). Of course being a conscientious system admin, it didn't stop me double checking them against the Oracle documentation each time in case some advice had changed.
This didn't happen by chance. But it meant that you could be reasonable certain no obscure kernel settings were incorrectly set (at least not by an oversight, didn't stop people setting the wrong settings when tuning).
At the time Oracle were talking with Hewlett Packard about a stripped down HP-UX to build "Oracle Servers" on PA-RISC. It made sense then, and it still makes sense, except HP-UX is no longer the "obvious choice" for an Oracle server.
To be honest, I think in the GNU/Linux world, it is choice of certified hardware that is probably as important, if not more so, for Oracle, than choice of distribution. Since I've been bitten by underdocumented, under tested, RAID hardware or Linux drivers for same (the effect is the same, no matter where the fault lies). If you are aiming for really high availability on an Oracle database, buying the solution as one stop from Oracle makes sense.
I doubt cost-wise it would be that competitive with DELL and Redhat, at least initially, but for some applications hardware cost is irrelevant compared to unplanned downtime.
Something like Debian, or Ubuntu, with long support periods, and completely freely redistributable base (with builtin rebranding -- "no Mozilla says you can't call it..." hazzles), is the obvious sort of base. Although presumably BSDs might be an option as well. Or Oracle might still want a big corporate backer for their distro variant.
> If I'm not mistake, you've spent a great deal of time with UserLinux and its predecessors in an attempt to solve the matter of a standardized platform.
/usr/share/doc/.
/etc/inetd.conf. So one uses dependencies when they save you redoing work, but design systems so that components don't interfere with each other unless the interaction is planned.
There are reasons to standardise components that aren't related to package management. Such as end user, and system admin training.
My understanding is that this was where UserLinux was coming from. The people involved all seemed very happy with how Debian packaging worked, but concerned they couldn't possibly support customers using; "KDE" and "Gnome"; Perl and Python; Ruby on Rails, and Catalyst; Postfix, and Exim.
Of course I switch between Debian and Windows, between Perl and PHP and ASP, between sendmail, and Postfix all day long, and as a result I'm pretty useless at all of them, and half the time can't remember what the comment character for the language or format I'm working in now is (I must learn Emacs, it handles that for you).
UserLinux basically built a list of preferred applications (a meta package), and let the Debian packaging system pull in the stuff it needed to make the list work. Thus using the power of the packaging system, not trying to work around it.
The packaging complexity problem is not as bad as you make out, because a lot of activities by different packages are completely orthogonal. i.e. Debian developers know they won't clash on documentation file location if they get the package names unique, because it will go in
It is building this "orthogonal" nature into systems that makes them more robust and scalable. Which is why xinetd should be preferred over inetd, because you can configure it by adding a file, rather than fighting over entries in
Surely the ultimate point is you can have any filesystem that fits in the Microsoft shaped filesystem hole, with free software you can change the hole. Okay 99% of the time it isn't wise to change the hole, but when it is Linus/Theo will recognise it and take the patch.
Anyway the other response on knowing "how it works", is summed up nicely if you go read the progress of some of the ReiserFS on Windows projects.
Hehe, surely the main difference with free software is you can run a similar quality tool over the source code, and know if the programmers involved sorted these kind of defects, or even just look it up at Coverity.
So you know if the program you are using is better than average or not.
I was actually pretty encouraged, I have Linux servers, running Apache, and Postfix, with Postgres backends, and Perl front ends, with a little PHP. Authentication via PAM.
When I looked at my deployed base of software, I'm seeing a lot of code with zero defects detected remaining (okay, a lot of them probably haven't been upgraded for reasons of system stability, if these fixes didn't make Debian stable, and I suspect a lot won't because they aren't that important).
Of course I know they are low defect products, that is why I chose them (without Coverity's help I might add). This leads the main security issues being my programming skills in creating web applications, and utilities. You don't need to use units as big as "per Kloc" when measuring my defect rate.
We also run some servers with Windows, IIS, various COM objects, and I can only assess the quality by how the software behaves, i.e. bugs affecting functionality, on that basis the equivalent proprietary software sucks in comparison, but is very clearly getting better recently. Of course these numbers say nothing directly about functionality issues, but I think experience tells us there is some correlation, otherwise how did I end up with so many low defect applications?
As others observed the published data speaks volumes more than the commentary. Since code can't have negative defects.
What no "qmail" statistics? Given Wietse noted some of the Postfix "defects" were false positives, I suspect it would cast more darkness than light, but hey they should test it just so we have something else to bitch about next week, now we've covered vi versus emacs.
Of course Emacs biggest defect is I don't have time to relearn how to use it, and that sort of defect wasn't covered either.
I switched a PC from Windows 95 to Debian, when Windows 95 went out of support.
Okay I'm not a typical PC user, but then I don't test any of the free software I hack on under Microsoft Windows anymore because I don't have a Microsoft Windows licence, and I don't see the point of paying for one (heck I just threw out the 120 day evaluation CDs for Windows 2000 Advanced Server I got at the launch party). So there were consequences for other Windows users of my switch, be it ever so small and insignificant.
So I don't think it is totally far fetched, but I doubt the Debian mirrors will crumble under the load.
Support is needed because there is software in Windows 98, that has known critical level security issues (I think unfixed vulnerabilities triggered by merely viewing images), so if they use the Internet, eventually they'll have to do something, if only reinstall and be more careful what images they view.
The great thing about kit like this is you can teach them how to dismantle and rebuild them safely without risking anything valuable. Guarantee you'll have the interest of most of them if you dismantle it, and explain the parts, especially if you make them help, or do it themselves. For kids dismantling things that are "expensive" is forbidden fruit.
;)
PC technology has barely changed internally from a visual perspective, apart from SATA replacing the ISA bus, and the amount of fans and pins on the CPUs.
It won't teach them anything useful for when they leave school on specific hardware, but in the mean time they'll be able to repair their own (and parents) kit, learn how not to get electrocuted, how not to wipe out sensitive electronics with static, they can learn about why complex things are made from components, the importance of standards, what a transformer is, what fuses are for, what is in a computer. One can even bring in the topic of mass production, and why such complex devices can be produced so cheaply.
Who knows, if you have some decent kit, they might even learn what a Faraday cage is. Okay scrub that one
Finally they can learn about recycling, environmentally sensitive disposable, landfill, leaching of heavy metals, etc.
Do watch out for sharp edges -- but hey that is a lesson in itself, and they'll learn it quickly enough.
I think the left/right thing will have to go. Even some very aggressively pro private enterprise, western governments have managed to end up with "bigger government" at the end of their term in Office.
But then I've yet to deprogram myself from the propaganda about free enterprise, and revisit the facts. There are a lot of unhelpful memes out there -- and a surprising number of them seem to be tenents of someones economic theory, and all of them are a lot less 'obviously true' than that axiom about parallel lines not meeting.
Governments should regulate markets, and prevent monopolies, if they do that well we've little to fear from big corporations. Only the extremists don't think governments should regulate markets, that is not to say some extremists haven't been influential.
Be warned some of those 'Swedish webcam chicks' are "its".
http://85.235.16.145/view/index.shtml
"Harvard says that because MS has more market share, it will have more market share."
They should read Nietzsche;
"The wind in the valley and the opinions of the marketplace of today indicate nothing of that which is coming but only of that which has been."
Yes if business continues as normal, Microsoft will continue to have a dominant marketshare in the desktop operating system market. Anyone who thinks things will continue as normal hasn't learnt much from the history of the last hundred years.
The whole point is the technology is seen as disruptive, that means business isn't continuing as before. The current GNU/Linux desktop share is effectively as close to zero as makes no odds, but if you look at who are the several million Desktop GNU/Linux users you see it is a very odd bunch, including some big opinion formers. I'm not that optimistic for a GNU/Linux desktop -- in a sense it doesn't matter much to me, I have a perfectly good free software desktop operating system at home and work, if the market grows a little, especially server side, and so I get a bit better hardware support I'll be happier.
What will make GNU/Linux as a desktop, if it ever does, is countries adopting it, or schools adopting it. Probably the biggest issue is lack of GNU/Linux know-how. People just don't know how to deploy it, or how even to set up basic desktop behaviours. Sure a few companies and individuals know, but it is probably the leading limiting factor for both business and home use, even over software availability and lock-in.
Absolutely disagree.
Best place to do spam filtering is at the termination point of the first SMTP connection. Okay in a perfect world this would be the end point, but the world isn't perfect.
User interaction is pretty much useless in spam filtering.
1) There is too much spam.
2) Users are too error prone (Trust me on this I have a SCOMP feed)
I have no choice I have to do spam filtering before the end point, because if I don't, the spam filtering at the end point will reject the email (and or blacklist my server), and it'll end up in my queue, and I'll have to try and return it (otherwise email would become unreliable), and some poor smuck gets 100,000 delivery reports for emails he never sent.
I have about 1000 domains on one of the email servers at work that forward the email elsewhere, and the average queue length is about 40 messages (with 5 day expiry). The queue consists almost entirely of spam from domains that won't accept email, refused by the final destination server. So I think I'm doing quite well on the spam filtering front.
We don't really do "false positives", I guess maybe a single report every few months, almost invariably due to misconfiguration of the sending email servers.
Read the article and learn.
As no one with Debian uses it, or looks in it, but XaraLX is in "non-free" for Etch and Sid. Looks like the maintainer upgraded it to 0.7 in Sid, Etch is 0.6 currently.
So that is why he is in Norway
The current schedule (AIUI) is for NetWare 6.5 to ship as a guest OS in Xen on SLES 10 mid-2007.
I assume the motivation is client driven. They may have ported a lot of their own code to GNU/Linux, but I expect there is a client base of installed Netware, with third party software etc. That said I haven't seen Netware running anywhere for years, but then I'm no longer consulting in a different place every week.
The folks at Novell have more motivation.
They have para-virtualisation of this thing called "Netware" running under SuSE (hmm sure I have a dusty certificate somewhere saying I'm certified on Netware). It lets Netware run on boxes that Netware doesn't have drivers for. It lets customers consolidate servers, upgrade hardware, and keep running their investment in Netware, and I bet Netware is a lot simpler to get running reliably (well as reliably as Netware ever runs, I wonder if SFT works under para-virtualisation) under SuSE, than say a whole enterprise GNU/Linux distro.
Redhat spent "millions" testing Xen ?! Seems a bit much given how much goes into testing some kernel changes.
Mod Matt down, his Karma is too good.
No one mentioned the deliberate mistake in the article, Ubuntu have to distribute the source for 3 years after they stop distributing the binaries, not "for as long as they distribute binaries".
I agree with the "it isn't difficult line". My Desktop PC has enough disk space to store all of the Debian mirrors and archives (and I'm not planning on doing a distribution with three kernels and 11 architectures), I might have to switch off mirroring, but the disks weren't the expensive bit of my PC. Since you can charge costs, all that matters is storage and organisation, storage is cheap, and storing the source code for stuff you distribute should be second nature for anyone with experience of source code management.
"DRM flight" (or perhaps we should label it the "freedom drain") is even better news, because that applies to Microsoft Vista as well.
/ has just switched to Debian Etch, and says he thinks it is even better than Ubuntu. But hey they are both free, he has a choice, although it did take him some effort to get all his weird Apple formatted music to play. I suspect on that score he switched "just in time".
Matt who did this Ubuntu on PPC thing earlier in the year http://mattl.co.uk/blog/2006/01/26/switcher-tales
He's got Vista on his other PC -- it is still booting.