That they added essential features is great, but that they still think they're unnecessary shows they have NO CLUE what challenges you get in big developments.
Who talk about "big developments" ? MySQL feature set is just fine for "small development", which happen to encompass most database deployement in use today. I am glad for you that Oracle have stored procedures and all that as you need these features, but realise most people don't.
Well, I would love to think dynamic partionning is useful. However, I am afraid it is priced out of reach. Instead of buying an expensive and proprieatry servers, why not just buy over-provisionned run-of-the-mill x86 servers ? Say, 20 quad Xeon Dell machines ?
Please correct me if I am wrong. These are just opinions, I do not know the actual price of these Unisys "mainframe", so I might be totally wrong on the price/performance ratio. I just think it is impossible for proprietary hardware to match the ratio of commidity x86 stuff.
I know people who tried Linux on S/390 and they say that, while VM is some seriously cool stuff, in the end it is just cheaper to buy a lot of x86 boxes (even when factoring in the added footprint in the data center).
I am working on the same problem, and it's not exactly a walk in the part. It's a big frigging deal actually. Lots and lots of stuff isn't going well.
What is so hard ? Copying users and groups ? Just copy/etc/passwd, group, shadow and gshadow (if your system files are from the pre-shadow era, migration tools exist). Preserving permission on files ? Mount old dir via NFS on the new box, and just copy them. Email ? Copy/var/spool/mail to your new box. What else ? Biggest problem I envision is how I will migrate the old printcap to a new CUPS-based machine. Hardly an insoluble problem.
That version of Windows was released in 1994. That was the same year that the Linux kernel 1.0 was released.
Maybe NT 4 existed when RH 5.1 was out, but it must have been pretty early in it's lifecycle.
However, I've actually upgraded an NT 3.5 system to Windows 2000, and it went very smoothly. Active Directory properly migrated every user exactly right, passwords, e-mail, the whole package. Surpriningly smooth.
Effectively, the easy part of migrating is being covered. Users, password and email is easy to migrate in Linux too, see above. How about the hard part, like whatever application you where running on this server beside the cookie-cutter file and print service ?
Microsoft has platform stability. The systems I am installing today are designed to run for 5 years. What migration path will I have for White Box Linux in five years?
This is the only sensible argument you put forth so far. But there still are a few caveat. Do you have a written guarantee that MS will have a migration path for whatever Windows-du-jour you are installing now ? If not, why do you display so much faith ?
If you consider nsswitch.conf and PAM to be "obscure and mysterious utilities", you are a pretty poor Linux admin. Sorry for the ad-hom, but you can't blame a system for your own ignorance. I would not know where to start if I had to make a Windows network application authenticate against an AD; does that make the process "obscure and mysterious" ?
RHEL if you have the dough, White Box Linux otherwise. Interestingly enough, this is what I am currently working on for a client (migrating a RH 5.2 box to WBL).
I am not sure if you imply that this is a Linux-specific problem, but if you do you are wrong. What would you recommend upgrading NT 3.51 to ? Will the upgrade to Win2k3 be entirely seemless and a walk in the park ?
How are the finance of the foundation doing ? What have they done with the money ? How many people have they kept employed via the foundation ? Who are the most generous corporate donators (so I can give them my business back) ? Inquiring minds want to know!
We are really hot about Nagios here. A nice thing to have to keep check on a fleet of servers.
Personnally, I am really enthusiast about centralizing user info and various config via nss_ldap (just need to convince my co-worker now). Directory services are cool.
"Flexibility of the Unix hardware" ? *cough* *cough*.... you must be joking, right ? At least, later generation of RISC hardware use standard such as PCI, which somewhat mitigate the closedness of these platforms.
That remind me... back a few years ago somewhere I do not work anymore, one of the client had a faulty motherboard on an RS/6000. Amount of the repair bill from IBM ? 6000$ CDN (about 4000 $ USD). Yep, this is not a typo : six, followed by three zeros.
When people tout the reliability of RISC/Unix hardware (yeah, right), the question that come to mind is : is it really worth the pricetag ?
This post is symptomatic of all the reasons Gentoo is despised in a large segment of the Linux population. First, Gentoo users are overly evangelical to the point of being annoying. Gentoo is a nice toy, a real impressive hack but it is not the right tools for most situation. I have seen Gentoo advocate recommend it to complete newbie, saying "installing software is easy, you just type emerge blah blah blah...". Pathetic.
Second, the vocal evangelist portion of the Gentoo community seem to be mostly beginner who just feel so empowered to be able to compile their own software. When you have been sysadmining professionnally for a while, compiling is not fun anymore, it become a chore you try to avoid. Binary packages, for all their imperfection, are convenient and predictable. And if you raise me an "emerge", I'll raise you an "apt", period.
If you think RedHat/Fedora users still deal with "RPM hell", you are sadly mistaken and out-of-date. While you recompile your software for the latest patches, I update with yum/apt-get/up2date. Welcome to the 21st century.
You don't get it. Everybody know about testing and unstable (talk about a reassuring naming scheme !). The problem is that both of these is in a constant state of flux. They are not snapshot; package change version regularly. Ideally, most sysadmin prefer a stable platform where software stay at a well-known revision. The latest such snpashot from Debian is what ? 3 years old ?
Notice that Fedora will suffer from a similar problems in the near future. Considerng the very short support window (6 months), user will have to run forward and deploy new revision to get bug fixes, which will cause a lot of headache.
1 hour, 4 hours, 3 days... it's somewhat unimportant, yes. The problem is that most people, including myself, don't see the point of compiling every little piece of software you install on your server. The reputed gain in performance is not worth it; hardware is cheap, cheaper than sysadmin time. I know you may use pre-compiled package with Gentoo, but then what is the point ? You may as well use Debian or a RedHat deritative. And this is exactly what I do.
Regex as operator, IMHO, is God-send in Perl. Regex-as-object in Python is clunky (too many step to apply a regex), and PHP implementation as function is confusing.
foreach() is a close second. I think Java and C# now have this operator too (unsure about Java). If I where stuck using the repetitive and redundant "for (assignation; stop condition; incrementation) { block }" construct for simple list iteration, it would make me scream.
If you are familiar with Perl, I suggest you read "Advanced Perl Programming" by Sriram Srinivasan from O'Reilly. Reading this baffle me with the flexibility and ingenuity of this language. While I don't advocate using all these techniques (tie, closure, typeglob, etc) in production code for maintainability purpose, knowing your options is certainly a good thing.
Slashdot Mirror
Stop right here! How much is one of those 32 CPU "mainframe" worth ? Without answering this question, you are speculating just as much as I was.
Who talk about "big developments" ? MySQL feature set is just fine for "small development", which happen to encompass most database deployement in use today. I am glad for you that Oracle have stored procedures and all that as you need these features, but realise most people don't.
Well, I would love to think dynamic partionning is useful. However, I am afraid it is priced out of reach. Instead of buying an expensive and proprieatry servers, why not just buy over-provisionned run-of-the-mill x86 servers ? Say, 20 quad Xeon Dell machines ?
Please correct me if I am wrong. These are just opinions, I do not know the actual price of these Unisys "mainframe", so I might be totally wrong on the price/performance ratio. I just think it is impossible for proprietary hardware to match the ratio of commidity x86 stuff.
I know people who tried Linux on S/390 and they say that, while VM is some seriously cool stuff, in the end it is just cheaper to buy a lot of x86 boxes (even when factoring in the added footprint in the data center).
Since you want to analyze ambient sound, should'nt you be using a lossless codec instead of MP3 ?
There is no way you could have supported the parent post better than you just did. You proved his point without a doubt.
If you've got tracert, you are running the wrong OS.
What is so hard ? Copying users and groups ? Just copy /etc/passwd, group, shadow and gshadow (if your system files are from the pre-shadow era, migration tools exist). Preserving permission on files ? Mount old dir via NFS on the new box, and just copy them. Email ? Copy /var/spool/mail to your new box. What else ? Biggest problem I envision is how I will migrate the old printcap to a new CUPS-based machine. Hardly an insoluble problem.
Maybe NT 4 existed when RH 5.1 was out, but it must have been pretty early in it's lifecycle.
Effectively, the easy part of migrating is being covered. Users, password and email is easy to migrate in Linux too, see above. How about the hard part, like whatever application you where running on this server beside the cookie-cutter file and print service ?
This is the only sensible argument you put forth so far. But there still are a few caveat. Do you have a written guarantee that MS will have a migration path for whatever Windows-du-jour you are installing now ? If not, why do you display so much faith ?
If you consider nsswitch.conf and PAM to be "obscure and mysterious utilities", you are a pretty poor Linux admin. Sorry for the ad-hom, but you can't blame a system for your own ignorance. I would not know where to start if I had to make a Windows network application authenticate against an AD; does that make the process "obscure and mysterious" ?
RHEL if you have the dough, White Box Linux otherwise. Interestingly enough, this is what I am currently working on for a client (migrating a RH 5.2 box to WBL).
I am not sure if you imply that this is a Linux-specific problem, but if you do you are wrong. What would you recommend upgrading NT 3.51 to ? Will the upgrade to Win2k3 be entirely seemless and a walk in the park ?
*cough*Texas*cough*
Superior intelligence and wit ;)
Sorry, sir. I thought you where the guy who said that building rpm was a dependency hell. My bad.
How are the finance of the foundation doing ? What have they done with the money ? How many people have they kept employed via the foundation ? Who are the most generous corporate donators (so I can give them my business back) ? Inquiring minds want to know!
<em class="cheerleader"> Go Mozilla ! </em>
We are really hot about Nagios here. A nice thing to have to keep check on a fleet of servers.
Personnally, I am really enthusiast about centralizing user info and various config via nss_ldap (just need to convince my co-worker now). Directory services are cool.
So does rpmbuild ... what is your point ?
"Flexibility of the Unix hardware" ? *cough* *cough* .... you must be joking, right ? At least, later generation of RISC hardware use standard such as PCI, which somewhat mitigate the closedness of these platforms.
That remind me ... back a few years ago somewhere I do not work anymore, one of the client had a faulty motherboard on an RS/6000. Amount of the repair bill from IBM ? 6000$ CDN (about 4000 $ USD). Yep, this is not a typo : six, followed by three zeros.
When people tout the reliability of RISC/Unix hardware (yeah, right), the question that come to mind is : is it really worth the pricetag ?
Would you mind explaining what these issues are ? Do you mean you can dispense with installing development packages under Debian ?
This post is symptomatic of all the reasons Gentoo is despised in a large segment of the Linux population. First, Gentoo users are overly evangelical to the point of being annoying. Gentoo is a nice toy, a real impressive hack but it is not the right tools for most situation. I have seen Gentoo advocate recommend it to complete newbie, saying "installing software is easy, you just type emerge blah blah blah...". Pathetic.
Second, the vocal evangelist portion of the Gentoo community seem to be mostly beginner who just feel so empowered to be able to compile their own software. When you have been sysadmining professionnally for a while, compiling is not fun anymore, it become a chore you try to avoid. Binary packages, for all their imperfection, are convenient and predictable. And if you raise me an "emerge", I'll raise you an "apt", period.
If you think RedHat/Fedora users still deal with "RPM hell", you are sadly mistaken and out-of-date. While you recompile your software for the latest patches, I update with yum/apt-get/up2date. Welcome to the 21st century.
You don't have to specify which package to build ?
You don't get it. Everybody know about testing and unstable (talk about a reassuring naming scheme !). The problem is that both of these is in a constant state of flux. They are not snapshot; package change version regularly. Ideally, most sysadmin prefer a stable platform where software stay at a well-known revision. The latest such snpashot from Debian is what ? 3 years old ?
Notice that Fedora will suffer from a similar problems in the near future. Considerng the very short support window (6 months), user will have to run forward and deploy new revision to get bug fixes, which will cause a lot of headache.
Just curious, but what could be easier than "rpmbuild -bb /usr/src/redhat/SPECS/blah.spec" ?
1 hour, 4 hours, 3 days ... it's somewhat unimportant, yes. The problem is that most people, including myself, don't see the point of compiling every little piece of software you install on your server. The reputed gain in performance is not worth it; hardware is cheap, cheaper than sysadmin time. I know you may use pre-compiled package with Gentoo, but then what is the point ? You may as well use Debian or a RedHat deritative. And this is exactly what I do.
Regex as operator, IMHO, is God-send in Perl. Regex-as-object in Python is clunky (too many step to apply a regex), and PHP implementation as function is confusing.
foreach() is a close second. I think Java and C# now have this operator too (unsure about Java). If I where stuck using the repetitive and redundant "for (assignation; stop condition; incrementation) { block }" construct for simple list iteration, it would make me scream.
If you are familiar with Perl, I suggest you read "Advanced Perl Programming" by Sriram Srinivasan from O'Reilly. Reading this baffle me with the flexibility and ingenuity of this language. While I don't advocate using all these techniques (tie, closure, typeglob, etc) in production code for maintainability purpose, knowing your options is certainly a good thing.