Asbestos? WTF? Asbestos has never been used for insulating individule cables, though possibly cable ducts. Had your toxic cables been insulated with asbestos, they wouldnt have caught fire in the first place.
There is definitely an advantage to using the same OS on both the desktop and the server. But there are at least as many reasons not to.. The most important one being genetic diversity. Do you really want a given worm to be able to take down both your servers and your desktops? ?
The distinction between versions of Windows is a few registry settings, and you are forced to run a insane amount of stuff not strictly necessary for a given role. The GUI being the prime example: You are forced to run it no matter what. If it only consumed resources that would be bad enough, but it also has lots of extra things running that are opportunity for security holes. Integrating a web browser into the GUI which is integrated with the OS means a exploit for one can get everything. Besides, how often do you log into the console of your server? You should be able to administer it remotely.
With Linux (and UNIX in general), I think you have the best of both worlds. Given the UNIX philosophy of small, isolated, tools, a desktop and a server (or even different servers doing different things) can have radically different things running. There is enough the same that experience, training, skills are useful for both roles.
There are regular reports on the Horde/IMP lists that PHPs gettext component is broken. As the other poster says, this is some how thread related; PHP not being thread safe. There is at least one report of languages that change within a single page... But it seems that this only happens under heavy load.
It would be interesting to know if PHP 5 will be thread safe, and this usable in production with Apache 2.x
eDirectory runs on Windows, Netware, many distros of Linux. Groupwise, same deal. Damm near every product they ship (that isnt Netware) runs everwhere. And everything they ship WILL run on Linux come "Netware" 7.0.
OS? What are you talking about? Who cares about the OS?
Do you still help them when they come to your desk? If so, when someone comes up to you desk you should not look at them, extend your arm towards them, palm out, fingers out, and say "talk to the email", and otherwise ignore them.
My Commodore 64 with its BASIC interperter way baci in 1982 knew what I meant. My current computer is about a gabizillion times faster, and has more memory then all 64s ever built put together. So dont try to talk to me about efficency.
Within 3m of me, I have 3 computers and a pile of 3 floppy drives. None of the computers have a floppy drive installed.
I only ever use floppies if/when Im installing a new OS. Depending on the phase of the moon, one of those floppy drives works. Exactly which one is random. After trying to make a boot disk on one of a stack of 30 floppy disks, I usually (but not always) get one that works.
A couple of weeks ago, I happened to have a new server in here that I installed FC-1 on. It had a PXEable NIC on the board, so I tried using that. It took about 10 minutes of effort, but it just worked. Had I tried to boot of a floppy I would still be swearing at the insanely stupid technology that should die. Die. DIE!
InnoDB is a sepearate project from MySQL proper, yes. But it comes with MySQL. It is the default. You have to do nothing special to install it (well, maby if your compiling it). Unless you specifiy something else in your CREATE TABLE statements, you get it by default for new tables..It is a distinction with out a difference.
Did you read my previous post? All database vendors implement a crippled version of the spec. If all I ever use is Oracle, then I will write SQL that works only on Oracle. If all I ever use is MSSQLServer then all the SQL I write will work only on MSSQLServer. It doesnt really matter who is closest or who is furthest away from the spec; they are all different enough to be incompatable.
What MySQL lacks is specific advanced features. You can live with out them, building the functionality with simple features into your app. This should be forwards compatable with new versions of MySQL, and cross compatable with existing SQL servers. That it is not, is not because MySQL is paticularly wrong, but that they are all wrong.
The argument could be made that by not implementing many SQL features you implicitly force your uses to be aware of the state of incompatabilities. MySQL docs are very clear on where and how they diverge from the standard. They are very clear that they DO diverge from the standard.
MySQL with InnoDB tables is ACID compliant. InnoDB tables are the default with 4.x, which has been out for 2+ years.
There is a SQL standard. Exactly 0 databases conform to (any version of) the standad, implement all of the features, and dont have extensions. All vendors implement things slightly different, and have their own extensions. Some of these diversion from the standard be compatable with other vendors , but such comptability is luck.
Many feel that this is intentional by DB vendors for customer lock in.
While it may be theoritecly possible to build a cross vendor compatable DB, the tempation to use usefull extensions is great. It happens all the time.. Any non-trivial database / SQL program will not be cross vendor comptable. For a bunch of reasons, a lot of MySQL code has been written by people who are unaware of the state of DB incompatibilties, and thus make little effort to use portable code. Or some people just dont care about portability. It is not MySQLs fault that ignorant or apathetic users rely heavily on its extenstions/incompatabilities.
SSH server key fingerprinting also pervents "man-in-the-middle" attacks. But generaly speaking people dont check the fingerprint on initial connection, and if/when it changes, don't check either.
(As implemented) SSL takes that one step further.. Not only are you sure you are sending data to the same place as last week (the key diddnt change), you are sure that who you are sending it to is who you think you are sending it to (name verified by a "trusted" third party). SSL probes something: that they are who they say they are.
But as for providing a measure of quality, no. And I doubt providing such a measure was even consitered a/potential/ function of CA back when this stuff was being implemented.
Dear Julius,
In order to provide out customers with the highest quality computing experience, we have numerous partnerships, one of which is with Microsoft. In order to minimize per-computer costs, we have brokered a deal with Microsoft so that Windows comes installed on all our deesktop systems that leave our factory.
If you wish, we can not install Microsoft Windows on your new 10,000 Optiplex systems, however we will still be paying Microsoft for Windows on those systems, and passing that cost on to you. Thus the Web Price quoted to you stands as is.
Thank you for you interest in Dell Computers. Have a nice day!
Starting off, the GPL (and LGPL) specificly allows for non-GPLd software to link against GPL stuff and not be contaiminated. The GPL specificly excluded the product of things like compilers (or text editor's, I suppose) from being contaiminated. Calls to functions provided by a GPLd package are ok. Calls accross the network to a GPLd server (like NTP) are even more ok. Think about IE connecting to an apache server (well, not apache - it doesnt have a viril license, but YKWIM).
But you know that... The question is about ambiguities of the (l)GPL. Interesting question. Novell is entering into this "free software" stuff in good faith. That said, it is almost a given that for systems as numerous and complex as Novell has, and will be porting to Linux, that when done, some lawyer will find a problem somewhere. Will the GNU sue NOVL without sending a heads up message: "we think X is non-compliant"? No, of course not. The will be given the opportunity to fix the problem, and they will. Slashdot has reported on a handfull of cases where GNU has sent out simmilar messages, most of which were addresses satisfactiorly.
The point of most licenses is to ensure that someone gets paid. If a violator is found some companies will be almost happy for the chance to sue someone and get paid fines and damages. Pay up front, or pay after litigation, the company wins either way. GNU, OTOH, has licenses for very different reasons. They dont want damages awarded to them. They want you to comply with the license. If threats bring a violator into compliance, then the GNUs interest has been served.
Secondly, your asking questions of what Novell will do in the future. They have already ported NDS to Linux - I had it running on a test server in 2000.
Very few OS companies produce compilers. Actually, only one OS company produces a compiler, Microsoft. All the other compilers that come with OSs come from hardware vendors. Sun doesnt ship Forte because it works well with Solaris, they ship it because it works well with Sparcs. The ship both Forte and Solaris so their hardware is usefull.
As for your second to last question, it answeres itself. Novell has the full source to NDS, and its 'native' platform, Netware. If there was a risk from GPL contamination (which there isn't) then they could just port over all those bits from Netware too.
It is unlikely that Novell will OSS Netware. It is likely so ridden with cross licensed technology that it would be all but impossible to do so. At the same time, if Netware has some nifty feature usefull to programers (us CN?s are fully aware of its nifty features for admins) Im sure there will be patches submitted to the approiate project - the kernel, glibc, gcc, or any one of dozens of others. Oracle, for example, submits lots of patches to lots of projects.
Im in a similar boat to you, though with perhaps a little more profesional experience. I can't claim to have gotten through the documentation though.
Security is directly related to the skill of the admin implementing it. The skill of an admin is directly related to how well that admin understands that tool. Not necessaraly the actual protocols and server bits that make it work, but at least its configuration. My point in experiementing was not to get a single link up, but to eventualy use it for securing WiFi users. Per customer config? No way. The config should be:
RADIUS server: x.x.x.x
RADIUS shared secret: xxxxxx
IP Range: x.x.x.x-x.x.x.x
I don't understand why it can't be that simple. Even if I was prepared to invest a few days in getting it working, it would be all but impossible to actualy USE it, getting normal people to also invest a few days in learning it. Thus, me spending a few days learning it would have been useless as it would never be used.
Their documentation is some of the worst I have ever seen for a project of its size. left/right; WTF? client/server.
It is unclear to me the distinction beteween freeswan.org, freeswan.ca, and superfreeswan, and what (if any) downside there is to using the only usefull one, superfreeswan. While there were hundreds to choose from, no prebuilt RPMs for my kernel version, and the srpm I happend to find via google produced a kernel module that had symbol mismatches.
This is like passwords. If you force people to choose long passwords with weird characters in them, they WILL write them down. The solution, if your realy concerned with password-level security, is to change to something else. Biometrics; smart cards, whatever. If security is too hard to use, people will circumvent it. Freeswan is just too fucking hard to use, so I won't use it at all.
Slashdot Mirror
You are allowed to make up your own words.
Sun spent $20million, bought a half ass competetor to MS Office and released it as OSS. Big deal. They spend that much each year on paperclips.
OOo is smoke and mirrors. It does not demonstrate that Sun is at all committed to OSS. It only demonstrates that Sun is committed to anti-Microsoft.
The distinction between versions of Windows is a few registry settings, and you are forced to run a insane amount of stuff not strictly necessary for a given role. The GUI being the prime example: You are forced to run it no matter what. If it only consumed resources that would be bad enough, but it also has lots of extra things running that are opportunity for security holes. Integrating a web browser into the GUI which is integrated with the OS means a exploit for one can get everything. Besides, how often do you log into the console of your server? You should be able to administer it remotely.
With Linux (and UNIX in general), I think you have the best of both worlds. Given the UNIX philosophy of small, isolated, tools, a desktop and a server (or even different servers doing different things) can have radically different things running. There is enough the same that experience, training, skills are useful for both roles.
It would be interesting to know if PHP 5 will be thread safe, and this usable in production with Apache 2.x
Boycot Microsoft search engine. Set your web server to refuse to talk to their spider.
Except that everything starts out as Linux, and Darl walks around taking them down. Hmm.. The Anti-XBill.
Cost: $2.9 billion
Life expectency: 5 years
And the robot control software runs on Linux.
Providing Anaconda with a kickstart config file is not quite the same thing as building your own distro.
OS? What are you talking about? Who cares about the OS?
Do you still help them when they come to your desk? If so, when someone comes up to you desk you should not look at them, extend your arm towards them, palm out, fingers out, and say "talk to the email", and otherwise ignore them.
My Commodore 64 with its BASIC interperter way baci in 1982 knew what I meant. My current computer is about a gabizillion times faster, and has more memory then all 64s ever built put together. So dont try to talk to me about efficency.
I only ever use floppies if/when Im installing a new OS. Depending on the phase of the moon, one of those floppy drives works. Exactly which one is random. After trying to make a boot disk on one of a stack of 30 floppy disks, I usually (but not always) get one that works.
A couple of weeks ago, I happened to have a new server in here that I installed FC-1 on. It had a PXEable NIC on the board, so I tried using that. It took about 10 minutes of effort, but it just worked. Had I tried to boot of a floppy I would still be swearing at the insanely stupid technology that should die. Die. DIE!
He said 8 digit SALES per year.
Things that happen every day are not news.
Did you read my previous post? All database vendors implement a crippled version of the spec. If all I ever use is Oracle, then I will write SQL that works only on Oracle. If all I ever use is MSSQLServer then all the SQL I write will work only on MSSQLServer. It doesnt really matter who is closest or who is furthest away from the spec; they are all different enough to be incompatable.
What MySQL lacks is specific advanced features. You can live with out them, building the functionality with simple features into your app. This should be forwards compatable with new versions of MySQL, and cross compatable with existing SQL servers. That it is not, is not because MySQL is paticularly wrong, but that they are all wrong.
The argument could be made that by not implementing many SQL features you implicitly force your uses to be aware of the state of incompatabilities. MySQL docs are very clear on where and how they diverge from the standard. They are very clear that they DO diverge from the standard.
There is a SQL standard. Exactly 0 databases conform to (any version of) the standad, implement all of the features, and dont have extensions. All vendors implement things slightly different, and have their own extensions. Some of these diversion from the standard be compatable with other vendors , but such comptability is luck.
Many feel that this is intentional by DB vendors for customer lock in.
While it may be theoritecly possible to build a cross vendor compatable DB, the tempation to use usefull extensions is great. It happens all the time.. Any non-trivial database / SQL program will not be cross vendor comptable. For a bunch of reasons, a lot of MySQL code has been written by people who are unaware of the state of DB incompatibilties, and thus make little effort to use portable code. Or some people just dont care about portability. It is not MySQLs fault that ignorant or apathetic users rely heavily on its extenstions/incompatabilities.
(As implemented) SSL takes that one step further.. Not only are you sure you are sending data to the same place as last week (the key diddnt change), you are sure that who you are sending it to is who you think you are sending it to (name verified by a "trusted" third party). SSL probes something: that they are who they say they are.
But as for providing a measure of quality, no. And I doubt providing such a measure was even consitered a /potential/ function of CA back when this stuff was being implemented.
In order to provide out customers with the highest quality computing experience, we have numerous partnerships, one of which is with Microsoft. In order to minimize per-computer costs, we have brokered a deal with Microsoft so that Windows comes installed on all our deesktop systems that leave our factory.
If you wish, we can not install Microsoft Windows on your new 10,000 Optiplex systems, however we will still be paying Microsoft for Windows on those systems, and passing that cost on to you. Thus the Web Price quoted to you stands as is.
Thank you for you interest in Dell Computers. Have a nice day!
It is impossible to purchase a desktop PC from name brand PC makers without Windows on it.
But you know that... The question is about ambiguities of the (l)GPL. Interesting question. Novell is entering into this "free software" stuff in good faith. That said, it is almost a given that for systems as numerous and complex as Novell has, and will be porting to Linux, that when done, some lawyer will find a problem somewhere. Will the GNU sue NOVL without sending a heads up message: "we think X is non-compliant"? No, of course not. The will be given the opportunity to fix the problem, and they will. Slashdot has reported on a handfull of cases where GNU has sent out simmilar messages, most of which were addresses satisfactiorly.
The point of most licenses is to ensure that someone gets paid. If a violator is found some companies will be almost happy for the chance to sue someone and get paid fines and damages. Pay up front, or pay after litigation, the company wins either way. GNU, OTOH, has licenses for very different reasons. They dont want damages awarded to them. They want you to comply with the license. If threats bring a violator into compliance, then the GNUs interest has been served.
Secondly, your asking questions of what Novell will do in the future. They have already ported NDS to Linux - I had it running on a test server in 2000.
Very few OS companies produce compilers. Actually, only one OS company produces a compiler, Microsoft. All the other compilers that come with OSs come from hardware vendors. Sun doesnt ship Forte because it works well with Solaris, they ship it because it works well with Sparcs. The ship both Forte and Solaris so their hardware is usefull.
As for your second to last question, it answeres itself. Novell has the full source to NDS, and its 'native' platform, Netware. If there was a risk from GPL contamination (which there isn't) then they could just port over all those bits from Netware too.
It is unlikely that Novell will OSS Netware. It is likely so ridden with cross licensed technology that it would be all but impossible to do so. At the same time, if Netware has some nifty feature usefull to programers (us CN?s are fully aware of its nifty features for admins) Im sure there will be patches submitted to the approiate project - the kernel, glibc, gcc, or any one of dozens of others. Oracle, for example, submits lots of patches to lots of projects.
Security is directly related to the skill of the admin implementing it. The skill of an admin is directly related to how well that admin understands that tool. Not necessaraly the actual protocols and server bits that make it work, but at least its configuration. My point in experiementing was not to get a single link up, but to eventualy use it for securing WiFi users. Per customer config? No way. The config should be:
RADIUS server: x.x.x.x
RADIUS shared secret: xxxxxx
IP Range: x.x.x.x-x.x.x.x
I don't understand why it can't be that simple. Even if I was prepared to invest a few days in getting it working, it would be all but impossible to actualy USE it, getting normal people to also invest a few days in learning it. Thus, me spending a few days learning it would have been useless as it would never be used.
Their documentation is some of the worst I have ever seen for a project of its size. left/right; WTF? client/server.
It is unclear to me the distinction beteween freeswan.org, freeswan.ca, and superfreeswan, and what (if any) downside there is to using the only usefull one, superfreeswan. While there were hundreds to choose from, no prebuilt RPMs for my kernel version, and the srpm I happend to find via google produced a kernel module that had symbol mismatches.
This is like passwords. If you force people to choose long passwords with weird characters in them, they WILL write them down. The solution, if your realy concerned with password-level security, is to change to something else. Biometrics; smart cards, whatever. If security is too hard to use, people will circumvent it. Freeswan is just too fucking hard to use, so I won't use it at all.
Sun incorporated in 1982. IBM PCs came out in 1981. Sun may be influential, but that influence has never been sufficent to effect things in the past.
I beleive (but am not sure) that this was an official 'sugestion' from IBM when the PC first came out, based on long established convention.