Large organizations tend to carry along overcompensated freeloaders. (Read CEO, CFO, etc.)
Do you have any supporting arguments/ Yes, CEOs make poor engineers. There's a culture within Slashdot that assumes if someone doesn't have anythign supremely technical to offer they are useless. But experience at a lot of tech companies shows most engineeers also make extremely poor CEOs.
Add apt-get support to kpackage for 3.1, and when it comes time to install 3.11, or 3.2, or any of the apps from apps.kde.org, you'll appreciate its advantages.
Nobody starts their day expecting to use an Operating System. They expect to do work, and the most bearing on how they do that work is the application they use to do it.
Why can't I be able to install a piece of software if it doesn't come with my system? Windows can do it, Linux should be able to as well. And it can: I do it all the time, on about 25 Linux machines.
The moral of the story isn't `don't ever install anything but your distro' the moral is `we need an easier way to install packages on most Linux distros'. On RH, usign apt-get from www.freshrpms.net serves this purpose well, and if the RH KDE packages were avaliable from such a source there would be few problems with installation.
Put the RPMs in an apt repository, make it avaliable by http, and run `apt-get install kdebase' on all your machines. Dependencies are automatically resolved as necessary to install the package. I do this for about 25 Linux workstations, all off one repository.
There is never any reason, ever, to use --nodeps. Luckily apt-get has `apt-get install -f' which performs a `fix' install to correct this kind of bad administration.
Damn straight. KDE could do a lot for its users by adding apt-get for RPM support to KPackage. Debian's nice, but there's a lot more Red Hat users as well as many other major distro's that are more popular, and most of these use the standard packaging format RPM (currently 3.0 is standard, 4.0 is likely to be when Maximum RPM is updated, which is likely later this year).
Already RH users are starting to get a lot of software avaliable via APT-get, including all of RH install CDs, the excellent Freshrpms archive (everything you wish you had but didn't) and Havoc Pennington's Gnomehide. Having this available through kpackage (rather than the apt-get command line, or an ugly tool like Synaptic) and creating APT archioves for KDE (I have one for my workplace - they're not difficult to create) would significantly enhance the install process.
On some of the Red Hat 7.2 boxes I administer, I run a nightly apt-get update and then apt-get update -s, which performs a simulated install and then mails me the output.
This way, I know when a new Red Hat update comes out, and have some idea about how successfully it will install on the system (generally OK, using Red Hat main, updates and freshrpms in my sources.list).
If you run Red Hat and you haven't discovered the joys of installing RPMs with APT, then I suggest you try it. You'll wonder how you ever got on without it.
Pointless waste of time. Better -- document every existing useful key in the default Windows registry or added by Microsoft software.
Well part of that, obviously, would be documenting which one's are useless and which ones aren't. Hence every registry key.
I'd allow all changes to a Windows server to be reversable and changable in time
Done
Not reliably enough that you could easily revery the registry and partition contents to a particular point in time without an EMC box.
Already most software does not require a reboot, even if it says it does. Exceptions would be anything that would modify the OS itself, and thus needs to touch system files that are currently in use (say, Service Packs and Hotfixes).
Exactly. MSDN says you don't need to reboot (well written apps using previous versions of the DLLS will continue to do so until they uload the library), but Microsoft's apps always want to. its hilariously bad that service packs and hotfixes require someone to schedule downtime for the system (and likely the users as well in a non clustered environment). Its one of the reasons Windows isn't regarded as a server OS by many used to Unixlike OSs where this is not the case.
Most software can already install in a silent mode (made easier if the software uses Windows Installer rather than some hacked-together method like NSIS). I don't think it's a requirement of the logo program, though.
Indeed. So have part of the logo program specify it. That way people would care again about whether a piece of software had the logo.
This is a matter of learning what's already there. Windows already has a very robust and powerful scripting engine. You just don't realize this because you're looking for something unix-like.
I disagree. I'm aware of VB / DCOM, and also aware of the improvements that MS is adding to their cli (netsh is a godsend). But there's still much work to be done in terms of making a simple, easy to use and highly powerful scripting language as the unix shell, and there are still l parts of Windows that cannot be called via COM and other mechanisms.
Windows is not Unix. Just because you're a successful Unix admin does not mean that translates to being a successful NT admin, and vice versa. Learn the system. That said, Microsoft has Services For Unix available, and there's always Cygwin.
I'm aware that Windows isn't Unix. But just as Linux keeps borrowing ideas from Windows, Windows keeps borrowing ideas from Unix (Win2ks storage system is rather Unixlike in nature, for example). Being able to play directly with devices that way the Systernals apps do would be handy, as would including OpenSSH in Windows (yes I know there's RDP, but yes I've tried to use it over a slow link).
And then you'd be screaming bloody murder that Microsoft is trying to force its way into the anti-virus industry.
I should clarify - most Antivirus vendors who had Exchange 5.5 solutions had to race the users for access to their mail, as 5.5 had now native antivirus API for AV vendors to use (Sybari famously avoided the problem by switching their own DLLs and Exchanges in the background, with obvious reaction from MS). They added somethinking like this to Exchange 2000, but its still nto quite stable with added features and fixes in each new E2000 SP. I'd just like it to be done already.
Of course, I didn't reply to all your points. Some of them were pretty good. These are just the ones that I had issues with.
I would document every existing key in the default Windows registry in a publically avaliable format, included with the OS.
I'd allow all changes to a Windows server to be reversable and changable in time
I would perform every step in MS lockdown guides as aprt of the default Windows install
I would make every Windows service run in a filesystem and process jail by their default install.
I wouldn't have any ports listening on a Windows server install by default (without a firewall).
I'd make part of the `Designed to Work With Blackcomb' badge program include a provision that to install or update a piece of software, a reboot must not be required.
Again, for the `Designed to Work With Blackcomb' badge program, I'd specifiy that software installation must not require interactive operation, allowing Windows admins to install thousands of packages simultaneously.
In the subsequent release of Windows, I'd provide that software providing service to end users must not need to restart or to accept changes.
I'd include SMS for free in Windows.
I would improve Windows scripting, and make command line interfaces available to the most popular Windows libraries, and libraries available for most popular Widows programs, so you could, for example, write a simple backup script and SMS the systems administrator if the backup fails
I'd include *all* necessary security updates on Windows update. I would sign up admins to recieve security notices as part of the post install wizard for Windows.
I would integrate the Event viewer and Windows messaging in some fashion to make sure administrative emails are being read.
I'd give Unix administrators the tools they need to work with and feel comfortable with my OS
I'd make a very scalable version of Exchange / Sharepoint that used a real database as a backend, that was trustworthy enough to be put directly on the net.
I'd allow servers running this protocol to connect with each other directly over the internet, bypassing SMTP, with many benefits to administrators for doing so. Thus I would push admins to remove SMTP based mailers from their MS based networks.
I'd include a virus scanning engine in Windows and Exchange that worked reliably
I'd pay external consultants to audit Windows code. Not excluding a large chunk of the OS like a certain BSD flavor, but the entire thing.
I'd make sure everyone on the planet knew at least five things you could do with Windows that you couldn't with Linux.
I'd create a new certification for Windows consultants with an emphasis on security and lab based, instructor graded tests.
I'd find out real world things that piss off Linux administrators about Linux (not things that Windows administrators are unsure about Linux). Then I'd make sure that Windows solved that problem.
I'd make sure people knew that we were making these changes in response to their demands, and because we're a competitive company, rather than a simple and bright but technically substanceless software company.
Most Linux users don't bother checking the crypto hashes on their downloadble binaries or reading the full sources of their application source. Creating an RPM (or dpkg, but RPM is both standard and more more widespread) virus would be one way to have viruses seriously make an impact on Linux users. Imagine all the APT repositories filled with corrupt rpms/dpkgs. Non foolingly, it's worth worrying about.
Fair enough - at least you've been a whole lot more civil than most of the replies.
With configure scripts, compiling is really easy. I am very incompetent, but I can compile GNOME from sources (not from GARNOME). Still, I've never managed to create binary packages for any distro (even though I've tried).
Keep trying - I'm not that skilled myself but creating packages is well within my reach. RPM (the standard Linux packaging system) has macros to handle any GNU autoconf/automake application, so most of what you have to do will be filling in specfiles.
I reckon there's a good chance you might not have found the right docs (because there's a lot of poor ones out there). Try freshrpms.net or IBM Developerworks for good packaging tutorials.
Ahem... "As well as not breaking your system, and ensuring a uninform install, uninstall / query process for all your software, your work is repeatable for other users and generally other distributions."
Who am I distributing it to?
Besides the abovementioned benefits, if you were a social, community minded sort of fellow (which I suggest from the subject line of your post you are not) then you might wish to help other users of your OS / distribution by distributing source / binary packages.
The moderator labeled you a troll because you said that people who felt uneasy in packaging and redistributing their system had no brains.
No I did not. If the moderator thought that, they didn't read what I wrote, which implied that if someone had the tchnical knowledge to compile source, then compiling packages was well within their abilities.
"Compiling Garnome on Sparc64 will take a similar amount of time than compiling it into packages"
Do you know this for sure? Have you tried to compile Gnome on Sparc64 yourself?
No. I have compiled other source applications into Solaris packages before. I see no reason why compiling Gnome would be any different. Feel free to provide me with one.
If it was so easy to compile Gnome onto a Solaris box
I never said it was easy to compile Gnome on a Solaris box.
Well, not entirely true. Ever try compiling Gnome for Sparc64? [grin] I wish somebody would make a package...
How is that not entirely true? I was responsing to the article where it suggests using compiling Garnome to install the latest GNOME beta, and suggesting compiling GNOME into packages (or both the source and binary packages). Compiling Garnome on Sparc64 will take a similar amount of time than compiling it into packages, and will provide a stable set of install metadata which can be used to install other packages on top of your GNOME, which seems likely.
I fail to understand how my post was in any way a troll.
You should make them - if you have the brains to compile software, you have the brains to package it. As well as not breaking your system, and ensuring a uninform install, uninstall / query process for all your software, your work is repeatable for other users and generally other distributions.
XUL is designed to be cross-platform and XML-based, not to be fast. That goal has been achieved, so XUL *IS* successfull.
I think XUL, which despite the processing power and RAM of modern machines still manages to perform than Windows 3.1 on a 486 is so unfeasibly slow it cancels any cross platform benefits, especially when QT already serves this purpose effectively. But Netscape users weren't screaming out for a cross platform GUI toolkit, they wanted a pleasant to use browser, which Mozilla, owing to XUL, is not.
if you haven't tried a nightly build recently, you aren't seeing the full picture.
I think I am. The reason for this is because people have been saying that Mozilla `is to slow' for a couple of years now, and other have been saying `if you haven't tried a nightly build recently, you aren't seeing the full picture' for an equal amount of time. I'm sorry, but Mozilla is too slow, and since its never been true in the past (otherwise why would so many people notice and complaqin) I see no reason why it would be in the future. I don't think you're telling the truth, and I don't think 1.0 will be faster than 0.99, no matter what you say.
Thank god for Galeon. The good stuff (gecko) with thefailed UI experiment (XUL).
Another item on my StarOffice "wish list" would be the ability to create and edit.pdf (Adobe Acrobat) files, something that is readily available for MS Word and Corel WordPerfect. Give me this feature, even if it's a plugin that costs $50 over and above whatever Sun decides to charge for StarOffice, and they'll get my money.
StarOffice 6 beta has this feature, and I'd be very surprised if the final didn't have it either. I think Roblimo just missed the menu option.
Slashdot Mirror
Paraphrased from The Official Ninja Homepage
"Source code" in an exe from a complete stranger! Let me rush out and get that!
So I take it you only ever install GPG signed source / binary packages? Or if installing unpackaged source review each line yourself?
Large organizations tend to carry along overcompensated freeloaders. (Read CEO, CFO, etc.)
Do you have any supporting arguments/ Yes, CEOs make poor engineers. There's a culture within Slashdot that assumes if someone doesn't have anythign supremely technical to offer they are useless. But experience at a lot of tech companies shows most engineeers also make extremely poor CEOs.
Its just you :).
Add apt-get support to kpackage for 3.1, and when it comes time to install 3.11, or 3.2, or any of the apps from apps.kde.org, you'll appreciate its advantages.
Nobody starts their day expecting to use an Operating System. They expect to do work, and the most bearing on how they do that work is the application they use to do it.
Why can't I be able to install a piece of software if it doesn't come with my system? Windows can do it, Linux should be able to as well. And it can: I do it all the time, on about 25 Linux machines.
The moral of the story isn't `don't ever install anything but your distro' the moral is `we need an easier way to install packages on most Linux distros'. On RH, usign apt-get from www.freshrpms.net serves this purpose well, and if the RH KDE packages were avaliable from such a source there would be few problems with installation.
rpm -Uvh http://enigma.freshrpms.net/pub/apt/apt-0.3.19cnc5 5-fr7.i386.rpm
Put the RPMs in an apt repository, make it avaliable by http, and run `apt-get install kdebase' on all your machines. Dependencies are automatically resolved as necessary to install the package. I do this for about 25 Linux workstations, all off one repository.
There is never any reason, ever, to use --nodeps. Luckily apt-get has `apt-get install -f' which performs a `fix' install to correct this kind of bad administration.
Too many reviews focus on installation.
Damn straight. KDE could do a lot for its users by adding apt-get for RPM support to KPackage. Debian's nice, but there's a lot more Red Hat users as well as many other major distro's that are more popular, and most of these use the standard packaging format RPM (currently 3.0 is standard, 4.0 is likely to be when Maximum RPM is updated, which is likely later this year).
Already RH users are starting to get a lot of software avaliable via APT-get, including all of RH install CDs, the excellent Freshrpms archive (everything you wish you had but didn't) and Havoc Pennington's Gnomehide. Having this available through kpackage (rather than the apt-get command line, or an ugly tool like Synaptic) and creating APT archioves for KDE (I have one for my workplace - they're not difficult to create) would significantly enhance the install process.
Mike
The first? Outlook and Outlook Express have been detonating them for years.
On some of the Red Hat 7.2 boxes I administer, I run a nightly apt-get update and then apt-get update -s, which performs a simulated install and then mails me the output.
This way, I know when a new Red Hat update comes out, and have some idea about how successfully it will install on the system (generally OK, using Red Hat main, updates and freshrpms in my sources.list).
If you run Red Hat and you haven't discovered the joys of installing RPMs with APT, then I suggest you try it. You'll wonder how you ever got on without it.
Pointless waste of time. Better -- document every existing useful key in the default Windows registry or added by Microsoft software.
Well part of that, obviously, would be documenting which one's are useless and which ones aren't. Hence every registry key.
I'd allow all changes to a Windows server to be reversable and changable in time
Done
Not reliably enough that you could easily revery the registry and partition contents to a particular point in time without an EMC box.
Already most software does not require a reboot, even if it says it does. Exceptions would be anything that would modify the OS itself, and thus needs to touch system files that are currently in use (say, Service Packs and Hotfixes).
Exactly. MSDN says you don't need to reboot (well written apps using previous versions of the DLLS will continue to do so until they uload the library), but Microsoft's apps always want to. its hilariously bad that service packs and hotfixes require someone to schedule downtime for the system (and likely the users as well in a non clustered environment). Its one of the reasons Windows isn't regarded as a server OS by many used to Unixlike OSs where this is not the case.
Most software can already install in a silent mode (made easier if the software uses Windows Installer rather than some hacked-together method like NSIS). I don't think it's a requirement of the logo program, though.
Indeed. So have part of the logo program specify it. That way people would care again about whether a piece of software had the logo.
This is a matter of learning what's already there. Windows already has a very robust and powerful scripting engine. You just don't realize this because you're looking for something unix-like.
I disagree. I'm aware of VB / DCOM, and also aware of the improvements that MS is adding to their cli (netsh is a godsend). But there's still much work to be done in terms of making a simple, easy to use and highly powerful scripting language as the unix shell, and there are still l parts of Windows that cannot be called via COM and other mechanisms.
Windows is not Unix. Just because you're a successful Unix admin does not mean that translates to being a successful NT admin, and vice versa. Learn the system. That said, Microsoft has Services For Unix available, and there's always Cygwin.
I'm aware that Windows isn't Unix. But just as Linux keeps borrowing ideas from Windows, Windows keeps borrowing ideas from Unix (Win2ks storage system is rather Unixlike in nature, for example). Being able to play directly with devices that way the Systernals apps do would be handy, as would including OpenSSH in Windows (yes I know there's RDP, but yes I've tried to use it over a slow link).
And then you'd be screaming bloody murder that Microsoft is trying to force its way into the anti-virus industry.
I should clarify - most Antivirus vendors who had Exchange 5.5 solutions had to race the users for access to their mail, as 5.5 had now native antivirus API for AV vendors to use (Sybari famously avoided the problem by switching their own DLLs and Exchanges in the background, with obvious reaction from MS). They added somethinking like this to Exchange 2000, but its still nto quite stable with added features and fixes in each new E2000 SP. I'd just like it to be done already.
Of course, I didn't reply to all your points. Some of them were pretty good. These are just the ones that I had issues with.
Thanks for your reply.
Mike
[ Reply to This | Parent ]
Most Linux users don't bother checking the crypto hashes on their downloadble binaries or reading the full sources of their application source. Creating an RPM (or dpkg, but RPM is both standard and more more widespread) virus would be one way to have viruses seriously make an impact on Linux users. Imagine all the APT repositories filled with corrupt rpms/dpkgs. Non foolingly, it's worth worrying about.
I don't really believe that.
Fair enough - at least you've been a whole lot more civil than most of the replies.
With configure scripts, compiling is really easy. I am very incompetent, but I can compile GNOME from sources (not from GARNOME). Still, I've never managed to create binary packages for any distro (even though I've tried).
Keep trying - I'm not that skilled myself but creating packages is well within my reach. RPM (the standard Linux packaging system) has macros to handle any GNU autoconf/automake application, so most of what you have to do will be filling in specfiles.
I reckon there's a good chance you might not have found the right docs (because there's a lot of poor ones out there). Try freshrpms.net or IBM Developerworks for good packaging tutorials.
Why would I want to package software?
Ahem...
"As well as not breaking your system, and ensuring a uninform install, uninstall / query process for all your software, your work is repeatable for other users and generally other distributions."
Who am I distributing it to?
Besides the abovementioned benefits, if you were a social, community minded sort of fellow (which I suggest from the subject line of your post you are not) then you might wish to help other users of your OS / distribution by distributing source / binary packages.
The moderator labeled you a troll because you said that people who felt uneasy in packaging and redistributing their system had no brains.
No I did not. If the moderator thought that, they didn't read what I wrote, which implied that if someone had the tchnical knowledge to compile source, then compiling packages was well within their abilities.
"Compiling Garnome on Sparc64 will take a similar amount of time than compiling it into packages"
Do you know this for sure? Have you tried to compile Gnome on Sparc64 yourself?
No. I have compiled other source applications into Solaris packages before. I see no reason why compiling Gnome would be any different. Feel free to provide me with one.
If it was so easy to compile Gnome onto a Solaris box
I never said it was easy to compile Gnome on a Solaris box.
Well, why don't you just show me how?
Sure. Visit freshrpms.net, rpm.org, or IBM developerworks for a couple of excellent tutorials on packaging.
Feel free to email me if you'd like to assist rather than talk shit.
I would, but you just insulted me, so now I rather wouldn't. People who abuse others like that give Linux a bad name.
Well, not entirely true. Ever try compiling Gnome for Sparc64? [grin] I wish somebody would make a package...
How is that not entirely true? I was responsing to the article where it suggests using compiling Garnome to install the latest GNOME beta, and suggesting compiling GNOME into packages (or both the source and binary packages). Compiling Garnome on Sparc64 will take a similar amount of time than compiling it into packages, and will provide a stable set of install metadata which can be used to install other packages on top of your GNOME, which seems likely.
I fail to understand how my post was in any way a troll.
You should make them - if you have the brains to compile software, you have the brains to package it. As well as not breaking your system, and ensuring a uninform install, uninstall / query process for all your software, your work is repeatable for other users and generally other distributions.
XUL is designed to be cross-platform and XML-based, not to be fast. That goal has been achieved, so XUL *IS* successfull.
I think XUL, which despite the processing power and RAM of modern machines still manages to perform than Windows 3.1 on a 486 is so unfeasibly slow it cancels any cross platform benefits, especially when QT already serves this purpose effectively. But Netscape users weren't screaming out for a cross platform GUI toolkit, they wanted a pleasant to use browser, which Mozilla, owing to XUL, is not.
Mozilla is a developpment framework. It's also a graphic toolkit, and a powerful language, whoose other components are based upon.
All Netscape users asked for was a good browser, soon.
if you haven't tried a nightly build recently, you aren't seeing the full picture.
I think I am. The reason for this is because people have been saying that Mozilla `is to slow' for a couple of years now, and other have been saying `if you haven't tried a nightly build recently, you aren't seeing the full picture' for an equal amount of time. I'm sorry, but Mozilla is too slow, and since its never been true in the past (otherwise why would so many people notice and complaqin) I see no reason why it would be in the future. I don't think you're telling the truth, and I don't think 1.0 will be faster than 0.99, no matter what you say.
Thank god for Galeon. The good stuff (gecko) with thefailed UI experiment (XUL).
This is off topic, but I really appreciate your sig, its very informative. Keep it around for a whle :)
As near as I can tell, the latest public build of OpenOffice still doesn't have this bug fixed.
Yes it has. From your link....
------- Additional Comments From mru@openoffice.org 2002-01-17 07:12 PST -------
Yes, works good in internal build 641c. Will reach OO with next public build.
Maybe two months ago, my laptop's win2k partition started getting scuzzy,
They do that? The partition or the disk? I've got a bunch of eyedeeyee disks lying around that do with getting scuzzy.
Another item on my StarOffice "wish list" would be the ability to create and edit .pdf (Adobe Acrobat) files, something that is readily available for MS Word and Corel WordPerfect. Give me this feature, even if it's a plugin that costs $50 over and above whatever Sun decides to charge for StarOffice, and they'll get my money.
StarOffice 6 beta has this feature, and I'd be very surprised if the final didn't have it either. I think Roblimo just missed the menu option.