The MySQL Administrator has nothing to do with phpMyAdmin. In fact it doesn't even have the same function. The MySQL Administrator is intended as a tool to administer the servers themselves, not the data that's in them (except to make backups, etc). This functionality will go a long ways to convincing PHB's who have a natural fear of anything that's not windows and not graphical that MySQL can be deployed and administered successfully within their enterprise. Everything from tuning the server engine itself to backups can be done from a comfortable console. Even to experienced Unix admins, this is a great tool.
I also agree. I have a basic linod virtual machine with 3 gb of disk space, soon to be bumped up to 4.5 gb. I've received immediate support help over IRC and can't be more pleased. You can install any distro you want practically on the linode virtual machine. Right now I'm running Fedora Core, because I really don't like debian. The terms of service are very good for most people's needs. In addition, freeswan support is in the linode kernel, as in tuntap support, so I can actually use freeswan or vpnc to connect to a vpn concentrator in my network at work if I needed to for whatever reason.
Having full root access (and therefore shell access) is very cool. This type of hosting beats all other types hands down, especially for the cost.
Having to wave a mouse cursor all over the screen to find out what things do is a bad UI decision in my opinion. OS X does the same thing for their window decorations, which I think is a bad idea, but they do use red,yellow, and green, which sort of helps you figure out what they might do.
UI elements should immediately convey their intended purpose and function (in most cases) rather than force the user to hilight them before telling exactly what it does. Purists would say right-clicking is bad for just that very reason. OS X has a happy medium here by showing a "gear" icon that you can always click on to get options for the current selected object, or you can right click if you know how.
Until recently, BYU had totally forgotten about their gopher server which was running without any changes for the last 7 years or so. gopher://gopher.byu.edu. I don't know if it is down now, or is just firewalled off.
Re:Really? Infamous?
on
Review: KDE 3.2
·
· Score: 3, Informative
Umm, the GNOME APIs are very well-designed and extremely object-oriented. As far as raw speed goes, gtk signal propagation is many times faster than QT's signals and slots mechanism.
Gnome/Gtk apps are not procedural. They are event-driven and object-oriented, just like QT. There are many apps written in the C++ bindings of Gnome. The idea that Gnome is fine for small programs but not large programs is false, if not stupid. Please do your research before posting such an obviously unresearched opinion.
As for overhead, C++ is, in many ways, simply syntactic sugar. As I said, there are awesome C++ bindings for Gnome and the Gnome object model (implemented in C) is very good. Programming overhead to do objects in C is minimal and there are great tools for generating objects. Also the switch to using XML primary to describe the guis makes Gnome GUI generation very easy and in just a few lines of code. In my experience your argument about reimplementing the OO is largely moot.
Also wxWindows is hardly a great example of an OO gui api. It's very much like MFC, which uses a event message mapping system, unlike QT and GTK which use signals and callbacks that can be dynamically created and conneccted together. I've never liked MFC, and I don't like wxWindows. I do like QT and GTK, though, particulary GTKMM.
Actually QEMU has mode where they are emulating the hardware. Right there are reports of windows 98 booting in the full virtual machine that qemu can now do. Obviously the linux binary loader part of qemu is faster and lighter, but qemu is a full emulator/virtual machine, soon to be on par with bochs and maybe virtualpc. There's even talk of using the qemu dynamic translation engine in bochs instead of the slow mechanism it currently uses.
Note that this darwine project is not a virtual machine per se, but a windows exe loader and api/ x86 code interpreter. The goal being to run windows exes on OS X as if they were native (this does not mean it has anything to do with quartz). In some respects, darwine could be compared best with the qemu linux loader portion of their project.
This is the fault of the application programmer. If a programmer knows how to write a decent message loop, you won't have this problem. Despite this, even poorly written program windows can be moved, they're just slow.
Umm, this is a design flaw in the Windows UI. No app should ever be able to interfere (even in a perceived way) with the operation of the system. I note that Window XP can move the windows around even when they are frozen, so obviously this is not an application-centric thing solely. Despite your claim, I have never seen a program window that was just slow to move. If it's frozen it was frozen.
But you are right. To make windows tolerable requires Mozilla Firebird, Thunderbird, OpenOffice, and the full cygwin unix toolset. Then I could run the gnome or kde desktop (full screen) and almost have a decent system.
Very true. And I'm not going to patch my servers yet. The problem is that OS X hardware is a bit pricey and I simply don't have any spares. I can take any old piece of crap machine and put linux on to test patches, however.
I'm not blaming Apple here. They are doing a good job trying to break into the server market and they have an excellent product, which I am quite happy with.
The kernel is the least of it all. The kernel is fairly transparent to a developer who knows darwin inside and out. When it comes to the kernel, linux for me is more transparent simply because I understand it better. I'm sure I will understand darwin better over time. But that's not what I was talking about.
The Opaqueness is in how everything is put together. Sure you can study darwin to figure it out. But the fact is that it's unix, but it's not unix. It's not system V, it has a hybrid init mechanism. Apple has also brought together many open source components, which is good, but it has done them in such a way that I can't just take the virgin code from, say, Samba, and compile. I can, however, get the code from apple. But now instead of being able to go to all the internet resources for help with a Samba 3.0 problem, I have to go to apple instead, since they have customized these components very heavily and the Samba developers can't make any real statement on a problem because fo that. It's just frustrating when there are problems. That's all. As with all proprietary operating systems, you really do tie yourself down to one vender. It's a calculated risk, one I'm not yet comfortable with (coming from an exclusive linux server setup) yet. Apple's tech support is very good, though. And the problems I've experienced will be resolved.
He's a special rep (sales and and tech) assigned to my University (and all the other universities in Utah). He's actually an Apple system engineer. He knows his stuff and has worked a lot with their OpenDirectory.
OS X in this regard is no better than Windows. It's an opaque operating system and dispite the list of changes that Apple provides, there's no real way to know if the patch is going to kill your system. For example, my local Apple rep warned me not to install the "12-19" security patch, as it will hose my AFP shares. Since this patch includes the 12-19 security patch, I doubt I'd install it without complete assurance from my apple rep. Furthermore, and update to the 10.3.2 service pack really borked my OpenDirectory (and from various web forums I learned that it's borked a lot of people's OpenDirectory). I am working with Apple to resolve this. However, I'm now as nervous as a Windows Sysadmin about installing this patch. Somehow I never felt this way, even with binary updates from RedHat. Linux is just so much more transparent and that makes me feel better than I do with these proprietary operating systems.
I have to run OS X. Linux is not well supported on the Powerbook 12". The combination of the unsupported Power Management Unit and the NVidia crappy Go5200 means that there's no sleep or any kind of power management for linux, which makes it not really useable as a laptop, sadly. Also Airport Extreme will not be supported (broadcom crap) anytime soon.
So I have to go the other way (linux on mac). OS X works very slickly in terms of the wireless and the sleep/suspend. Works so much more reliably than windows.
Developing for multiple platforms simultaneously. (OS X, linux, and windows via wine via the qemu x86 linux emulator). I already have compilers on OS X to target linux (x86 and ppc) and windows (mingw), but for the linux targets, I would love to be able to test binaries right there as I'm working.
Besides I just like Linux better (it's more transparent; OS X may be unix, but it's almost as opaque as Windows, no pun intended).
I currently do have the entire gnome 2.4 desktop up and running (on top of my quartz desktop) and that's nice. But there are just some things having linux around for would be nice.
To run coLinux right now, you still need the Cygwin XFree86 server. All coLinux can do is essentially a text-mode virtual console. This is okay, since running the X server as a native windows app is potentially faster. I frequently use Xnest in Linux as the display for my VMWare machines because it's so much faster. So there's no real need or hurry to get virtual graphics card support for coLinux, in my opinion.
According to the web site, the architecture of the software that makes this all possible is very portable and could be ported to Solaris, for example , allowing the running of Linux/Sparc on top of it, at full speed. I would love to see this ported to OS X. I love my powerbook and I like OS X, but running linux at the same time would be a huge benefit for me. I'll be following this project closely.
Emulation and virtualization are the coolest technologies I've ever seen.
The fd.o xserver is about 1 mb, and the X11 libraries to drive your apps weigh in at around 1 mb or so stripped, I think. At least on my fd.o installation they do. Also kdrive + libs + gtk2 + apps runs comfortably on a device like the Zaurus. Of course that's with no openGL stuff, or xinerama or xprint. But it does include freetype, xft, xrender, xdamage, composite, etc. The basic libraries are quite compact. If you really look at it, 1mb for kdrive, 1mb for libraries, plus maybe 5 mb for gtk2, an X11-based enbedded environment (supporting just one kind of display hardware) is very light and competitive with the Qtopia framebuffer system. Given that, I can't see any reason do use qt-embedded or gtk-fb for most things.
Some of those things in your list are not really libraries providing an api, but rather utilities, many of which on an embedded environment aren't needed.
I can't agree more. I recently wrote a fairly complicated proxy server and using glib (combined with the gnet libraries) has completely saved me. The glib routines for building quick hashtables, lists, and dynamic strings (all in C) make so many aspects of my life easier. By using a glib dynamic string as my input buffer, I can easily grow it to accomodate the incoming data rather than having to do all the realloc stuff myself.
I think glib (at least the routines for data types -- lists, hash tables, strings, etc) should be in the C standard library. The gobject stuff, while useful, should always be in a separate library.
Right now, I believe Keith is busy defining and prototyping the new extensions to the X protocol that will make our displays much faster and also prettier. Once the extensions are defined and tested on his kdrive architecture, then he'll probably take them to the this X group and work with them on their implementation.
Some of the things we'll likely see out of this: - the DAMAGE extension combined with double buffering for each window which eliminates almost all application redraws. - Smooth window resizing (no more jerky gtk/qt redraws when this happens), which will require widget support - Composite manager and alpha channel visual - allow true alpha for displaying of icons, translucency, shadows, etc without the hacks and will actually increase the perceived overal speed of the gui. - The adoption of Cairo by the main widget sets (gtk, qt) which will allow true vector-based widget and drawing toolkits which will look better and scale to higher resolutions better than even Quartz (which is still bitmap based)
It's not just about eye candy. It's about speed and flexibility. X is not going to be dead for many years, I think. Oh and by the way, with all these capabilities, kdrive is still under 1 MB and people are currently using it and the preliminary composite manager on iPAQs and Zauruses (see http://gpe.handhelds.org). On a small screen alpha blending the edges of windows and other ui elements makes things a lot more pleasant to deal with.
The future of X11 is very good and very bright. I'm excited to see what's coming. Cairo combined with composite and alpha extensions will bring us a truly awesome and scalable interface.
There is a command line utility that dumps the whole password database out (for backup purposes). You can use that to backup the password database plus openldap, and then reinitialize opendirectory under the new ip address and then reimport.
That said, you are correct about Linux. In many situations, it is the best way to go.
I Googled for it and came up with the answer: if you set up a password server on 10.2 Server you can't change the machine's IP address. This has to do with how Apple built the LDAP system that manages passwords, so it's a requirement if you want Windows file sharing (Samba in disguise) enabled. Which I did.
I called Apple tech support and they confirmed it: if I wanted to do Windows file sharing, I needed to set up a "password server" (LDAP). And if I set up a password server, I couldn't change the IP address of the machine. Ever. If I did, the users and groups would vanish into thin air. I asked if I could back up the user and group databases and then upload them again... and they said no. Not without stripping the passwords out. So I'd have to have my users reset their passwords.
Well, it's not like the ldap database disappears; it would still be on the disk. Probably the solution is to do a commandline dump of the whole OpenDirectory database (LDAP+Password server) that you can reimport after the change. We've used this method when I completely reinstalled the server after we found that 10.3.2 upgrade crippled our OpenDirectory install.
The Workgroup Manager (as mentioned by another poster) combined with OpenDirectory and Samba 3.0 make a killer competitor to MS's Active Directory. Definitely very cool and very easy to use and set up to server your windows needs. Panther Server is probably the best solution to bridge between your Windows, Apple, and Linux clients. Combine this with very good, cheap storage (2 TB fibre channel raid), and it's hard to beat.
While you can easily do everything that Panther Server does on your linux box, what Panther excels in is integration. Specifically the binding together of OpenLDAP, Samba, Apache, Postfix, IMAP, POP, and CUPS with the OpenDirectory password server. OpenDirectory's password server is essentially a SASL password store that they've hacked all the programs mentioned to interact directly with it for all authentication. Think of it similarly to what pam does for linux. The nice thing about OpenDirectory is that a password change from any of these mechanisms (say via samba) then all of the password hashes in the database are automatically synced (even kerberos is synced). This makes for very slick administration of users all from one central console. In the past on Linux, it was not uncommon to have to hack together some scripts to syncronize ldap, samba, and kerberos authentication stores. Even in the best case right now, samba password hashes have to be kept in the ldap database along with either an md5 hash for unix logins, or a pointer to kerberos. With OpenDirectory, there are no passwords stored in ldap itself. Instead an Apple Password field points to the password database which can provide md5 challenges and responses, samba challenges, and general password verification.
Essentially OpenDirectory brings all the technologies together that we already use and make them into a service that competes very well with Active Directory or NDS.
Another bonus is that since OpenDirectory (all its parts including the SASL password database and patches to cups, samba, etc) is open source, we could build a complete OpenDirectory-compatible system on Linux. I plan to do this over the next year or so. Most likely there will have to be a pam module created, and some patches made to OpenLDAP, Samba, etc. But it's a very exciting example of how to put open source projects together and have them work really well.
When Paramount Studios tried to cancel the original series back in the late 60's, Bjo Trimble organized star trek fans to undertake a huge letter-writing campaign that helped paramount decided to renew Star Trek's contract for another season. There was a tremendous groundswell of support among the fans at the time.
This time there will be no grassroots fan support. Unlike Star Trek, Enterprise has no real fan base. Enterprise is a shallow show, which doesn't delve into politics or social issues which made Star Trek so powerful. Exploring Vulcan sensuality is not a social issue.
This is in part a design decision, and partly due to the traditional nature of the gtk widget set. Essentially there is no MDI in gtk, for good or bad. Part of the solution to address this multiple window out-of-control problem is the dock idea that they've come up with. I configure gimp to just have one window (the default is 2, but you can easily rearrange it into one) that has the tool palette and all the other things I need docked to it. It's very dynamic and customizable. Sections can be added, tabs added, etc. That one window is always on top. Then my picture windows float next to it. That way I get the best of both worlds. I don't need another application to take over my entire screen like photoshop does. I prefer the windows. Except for the fact that the dock thing can only be vertical (having it horzonal at times would be nice), I think this is superior to the photoshop-style ui.
Besides that, I've always maintained that anyone who runs apps full screen (which you pretty much have to with MDI apps) really isn't using a windowing system to their best advantage. MDI is rapidly falling out of favor. MS no longer uses it for many applications and MacOS never ever did. Tabs work well for most things, although images are better off in windows. Anyway, the interface on gimp is light years ahead of the old interface! Now if only glade could get a similar interface makeover.
Slashdot Mirror
The MySQL Administrator has nothing to do with phpMyAdmin. In fact it doesn't even have the same function. The MySQL Administrator is intended as a tool to administer the servers themselves, not the data that's in them (except to make backups, etc). This functionality will go a long ways to convincing PHB's who have a natural fear of anything that's not windows and not graphical that MySQL can be deployed and administered successfully within their enterprise. Everything from tuning the server engine itself to backups can be done from a comfortable console. Even to experienced Unix admins, this is a great tool.
Let us know how your trial run goes.
I also agree. I have a basic linod virtual machine with 3 gb of disk space, soon to be bumped up to 4.5 gb. I've received immediate support help over IRC and can't be more pleased. You can install any distro you want practically on the linode virtual machine. Right now I'm running Fedora Core, because I really don't like debian. The terms of service are very good for most people's needs. In addition, freeswan support is in the linode kernel, as in tuntap support, so I can actually use freeswan or vpnc to connect to a vpn concentrator in my network at work if I needed to for whatever reason.
Having full root access (and therefore shell access) is very cool. This type of hosting beats all other types hands down, especially for the cost.
Having to wave a mouse cursor all over the screen to find out what things do is a bad UI decision in my opinion. OS X does the same thing for their window decorations, which I think is a bad idea, but they do use red,yellow, and green, which sort of helps you figure out what they might do.
UI elements should immediately convey their intended purpose and function (in most cases) rather than force the user to hilight them before telling exactly what it does. Purists would say right-clicking is bad for just that very reason. OS X has a happy medium here by showing a "gear" icon that you can always click on to get options for the current selected object, or you can right click if you know how.
Michael
Until recently, BYU had totally forgotten about their gopher server which was running without any changes for the last 7 years or so. gopher://gopher.byu.edu. I don't know if it is down now, or is just firewalled off.
Umm, the GNOME APIs are very well-designed and extremely object-oriented. As far as raw speed goes, gtk signal propagation is many times faster than QT's signals and slots mechanism.
Gnome/Gtk apps are not procedural. They are event-driven and object-oriented, just like QT. There are many apps written in the C++ bindings of Gnome. The idea that Gnome is fine for small programs but not large programs is false, if not stupid. Please do your research before posting such an obviously unresearched opinion.
As for overhead, C++ is, in many ways, simply syntactic sugar. As I said, there are awesome C++ bindings for Gnome and the Gnome object model (implemented in C) is very good. Programming overhead to do objects in C is minimal and there are great tools for generating objects. Also the switch to using XML primary to describe the guis makes Gnome GUI generation very easy and in just a few lines of code. In my experience your argument about reimplementing the OO is largely moot.
Also wxWindows is hardly a great example of an OO gui api. It's very much like MFC, which uses a event message mapping system, unlike QT and GTK which use signals and callbacks that can be dynamically created and conneccted together. I've never liked MFC, and I don't like wxWindows. I do like QT and GTK, though, particulary GTKMM.
Actually QEMU has mode where they are emulating the hardware. Right there are reports of windows 98 booting in the full virtual machine that qemu can now do. Obviously the linux binary loader part of qemu is faster and lighter, but qemu is a full emulator/virtual machine, soon to be on par with bochs and maybe virtualpc. There's even talk of using the qemu dynamic translation engine in bochs instead of the slow mechanism it currently uses.
Note that this darwine project is not a virtual machine per se, but a windows exe loader and api/ x86 code interpreter. The goal being to run windows exes on OS X as if they were native (this does not mean it has anything to do with quartz). In some respects, darwine could be compared best with the qemu linux loader portion of their project.
Of course Samba would work out of the box, but it would not work with OpenDirectory or the OpenLDAP schema that apple uses. Those require patches.
But you are right. To make windows tolerable requires Mozilla Firebird, Thunderbird, OpenOffice, and the full cygwin unix toolset. Then I could run the gnome or kde desktop (full screen) and almost have a decent system.
Very true. And I'm not going to patch my servers yet. The problem is that OS X hardware is a bit pricey and I simply don't have any spares. I can take any old piece of crap machine and put linux on to test patches, however.
I'm not blaming Apple here. They are doing a good job trying to break into the server market and they have an excellent product, which I am quite happy with.
The kernel is the least of it all. The kernel is fairly transparent to a developer who knows darwin inside and out. When it comes to the kernel, linux for me is more transparent simply because I understand it better. I'm sure I will understand darwin better over time. But that's not what I was talking about.
The Opaqueness is in how everything is put together. Sure you can study darwin to figure it out. But the fact is that it's unix, but it's not unix. It's not system V, it has a hybrid init mechanism. Apple has also brought together many open source components, which is good, but it has done them in such a way that I can't just take the virgin code from, say, Samba, and compile. I can, however, get the code from apple. But now instead of being able to go to all the internet resources for help with a Samba 3.0 problem, I have to go to apple instead, since they have customized these components very heavily and the Samba developers can't make any real statement on a problem because fo that. It's just frustrating when there are problems. That's all. As with all proprietary operating systems, you really do tie yourself down to one vender. It's a calculated risk, one I'm not yet comfortable with (coming from an exclusive linux server setup) yet. Apple's tech support is very good, though. And the problems I've experienced will be resolved.
He's a special rep (sales and and tech) assigned to my University (and all the other universities in Utah). He's actually an Apple system engineer. He knows his stuff and has worked a lot with their OpenDirectory.
OS X in this regard is no better than Windows. It's an opaque operating system and dispite the list of changes that Apple provides, there's no real way to know if the patch is going to kill your system. For example, my local Apple rep warned me not to install the "12-19" security patch, as it will hose my AFP shares. Since this patch includes the 12-19 security patch, I doubt I'd install it without complete assurance from my apple rep. Furthermore, and update to the 10.3.2 service pack really borked my OpenDirectory (and from various web forums I learned that it's borked a lot of people's OpenDirectory). I am working with Apple to resolve this. However, I'm now as nervous as a Windows Sysadmin about installing this patch. Somehow I never felt this way, even with binary updates from RedHat. Linux is just so much more transparent and that makes me feel better than I do with these proprietary operating systems.
I have to run OS X. Linux is not well supported on the Powerbook 12". The combination of the unsupported Power Management Unit and the NVidia crappy Go5200 means that there's no sleep or any kind of power management for linux, which makes it not really useable as a laptop, sadly. Also Airport Extreme will not be supported (broadcom crap) anytime soon.
So I have to go the other way (linux on mac). OS X works very slickly in terms of the wireless and the sleep/suspend. Works so much more reliably than windows.
Developing for multiple platforms simultaneously. (OS X, linux, and windows via wine via the qemu x86 linux emulator). I already have compilers on OS X to target linux (x86 and ppc) and windows (mingw), but for the linux targets, I would love to be able to test binaries right there as I'm working.
Besides I just like Linux better (it's more transparent; OS X may be unix, but it's almost as opaque as Windows, no pun intended).
I currently do have the entire gnome 2.4 desktop up and running (on top of my quartz desktop) and that's nice. But there are just some things having linux around for would be nice.
To run coLinux right now, you still need the Cygwin XFree86 server. All coLinux can do is essentially a text-mode virtual console. This is okay, since running the X server as a native windows app is potentially faster. I frequently use Xnest in Linux as the display for my VMWare machines because it's so much faster. So there's no real need or hurry to get virtual graphics card support for coLinux, in my opinion.
According to the web site, the architecture of the software that makes this all possible is very portable and could be ported to Solaris, for example , allowing the running of Linux/Sparc on top of it, at full speed. I would love to see this ported to OS X. I love my powerbook and I like OS X, but running linux at the same time would be a huge benefit for me. I'll be following this project closely.
Emulation and virtualization are the coolest technologies I've ever seen.
The fd.o xserver is about 1 mb, and the X11 libraries to drive your apps weigh in at around 1 mb or so stripped, I think. At least on my fd.o installation they do. Also kdrive + libs + gtk2 + apps runs comfortably on a device like the Zaurus. Of course that's with no openGL stuff, or xinerama or xprint. But it does include freetype, xft, xrender, xdamage, composite, etc. The basic libraries are quite compact. If you really look at it, 1mb for kdrive, 1mb for libraries, plus maybe 5 mb for gtk2, an X11-based enbedded environment (supporting just one kind of display hardware) is very light and competitive with the Qtopia framebuffer system. Given that, I can't see any reason do use qt-embedded or gtk-fb for most things.
Some of those things in your list are not really libraries providing an api, but rather utilities, many of which on an embedded environment aren't needed.
I can't agree more. I recently wrote a fairly complicated proxy server and using glib (combined with the gnet libraries) has completely saved me. The glib routines for building quick hashtables, lists, and dynamic strings (all in C) make so many aspects of my life easier. By using a glib dynamic string as my input buffer, I can easily grow it to accomodate the incoming data rather than having to do all the realloc stuff myself.
I think glib (at least the routines for data types -- lists, hash tables, strings, etc) should be in the C standard library. The gobject stuff, while useful, should always be in a separate library.
Right now, I believe Keith is busy defining and prototyping the new extensions to the X protocol that will make our displays much faster and also prettier. Once the extensions are defined and tested on his kdrive architecture, then he'll probably take them to the this X group and work with them on their implementation.
Some of the things we'll likely see out of this:
- the DAMAGE extension combined with double buffering for each window which eliminates almost all application redraws.
- Smooth window resizing (no more jerky gtk/qt redraws when this happens), which will require widget support
- Composite manager and alpha channel visual - allow true alpha for displaying of icons, translucency, shadows, etc without the hacks and will actually increase the perceived overal speed of the gui.
- The adoption of Cairo by the main widget sets (gtk, qt) which will allow true vector-based widget and drawing toolkits which will look better and scale to higher resolutions better than even Quartz (which is still bitmap based)
It's not just about eye candy. It's about speed and flexibility. X is not going to be dead for many years, I think. Oh and by the way, with all these capabilities, kdrive is still under 1 MB and people are currently using it and the preliminary composite manager on iPAQs and Zauruses (see http://gpe.handhelds.org). On a small screen alpha blending the edges of windows and other ui elements makes things a lot more pleasant to deal with.
The future of X11 is very good and very bright. I'm excited to see what's coming. Cairo combined with composite and alpha extensions will bring us a truly awesome and scalable interface.
There is a command line utility that dumps the whole password database out (for backup purposes). You can use that to backup the password database plus openldap, and then reinitialize opendirectory under the new ip address and then reimport.
That said, you are correct about Linux. In many situations, it is the best way to go.
The Workgroup Manager (as mentioned by another poster) combined with OpenDirectory and Samba 3.0 make a killer competitor to MS's Active Directory. Definitely very cool and very easy to use and set up to server your windows needs. Panther Server is probably the best solution to bridge between your Windows, Apple, and Linux clients. Combine this with very good, cheap storage (2 TB fibre channel raid), and it's hard to beat.
While you can easily do everything that Panther Server does on your linux box, what Panther excels in is integration. Specifically the binding together of OpenLDAP, Samba, Apache, Postfix, IMAP, POP, and CUPS with the OpenDirectory password server. OpenDirectory's password server is essentially a SASL password store that they've hacked all the programs mentioned to interact directly with it for all authentication. Think of it similarly to what pam does for linux. The nice thing about OpenDirectory is that a password change from any of these mechanisms (say via samba) then all of the password hashes in the database are automatically synced (even kerberos is synced). This makes for very slick administration of users all from one central console. In the past on Linux, it was not uncommon to have to hack together some scripts to syncronize ldap, samba, and kerberos authentication stores. Even in the best case right now, samba password hashes have to be kept in the ldap database along with either an md5 hash for unix logins, or a pointer to kerberos. With OpenDirectory, there are no passwords stored in ldap itself. Instead an Apple Password field points to the password database which can provide md5 challenges and responses, samba challenges, and general password verification.
Essentially OpenDirectory brings all the technologies together that we already use and make them into a service that competes very well with Active Directory or NDS.
Another bonus is that since OpenDirectory (all its parts including the SASL password database and patches to cups, samba, etc) is open source, we could build a complete OpenDirectory-compatible system on Linux. I plan to do this over the next year or so. Most likely there will have to be a pam module created, and some patches made to OpenLDAP, Samba, etc. But it's a very exciting example of how to put open source projects together and have them work really well.
When Paramount Studios tried to cancel the original series back in the late 60's, Bjo Trimble organized star trek fans to undertake a huge letter-writing campaign that helped paramount decided to renew Star Trek's contract for another season. There was a tremendous groundswell of support among the fans at the time.
This time there will be no grassroots fan support. Unlike Star Trek, Enterprise has no real fan base. Enterprise is a shallow show, which doesn't delve into politics or social issues which made Star Trek so powerful. Exploring Vulcan sensuality is not a social issue.
This is in part a design decision, and partly due to the traditional nature of the gtk widget set. Essentially there is no MDI in gtk, for good or bad. Part of the solution to address this multiple window out-of-control problem is the dock idea that they've come up with. I configure gimp to just have one window (the default is 2, but you can easily rearrange it into one) that has the tool palette and all the other things I need docked to it. It's very dynamic and customizable. Sections can be added, tabs added, etc. That one window is always on top. Then my picture windows float next to it. That way I get the best of both worlds. I don't need another application to take over my entire screen like photoshop does. I prefer the windows. Except for the fact that the dock thing can only be vertical (having it horzonal at times would be nice), I think this is superior to the photoshop-style ui.
Besides that, I've always maintained that anyone who runs apps full screen (which you pretty much have to with MDI apps) really isn't using a windowing system to their best advantage. MDI is rapidly falling out of favor. MS no longer uses it for many applications and MacOS never ever did. Tabs work well for most things, although images are better off in windows. Anyway, the interface on gimp is light years ahead of the old interface! Now if only glade could get a similar interface makeover.