I run k3b along with Evolution, gaim, openoffice, firefox, eclipse, vmware and xmms all on a gnome desktop. They all work well together, look good, are intuitive to use... the lone Piece of Shit I have installed is Acrobat Reader 5, which is a throwback to cheesy 1994 style X-Window interfaces. I'm so looking forward to the 7.0 beta.
So the one commercial application I have installed is also the most mundane, most non-standard widget wise. Funny how that works, huh?
In my 11 years of using X-Windows, middle click has NEVER failed to work with copy/paste.
Some people don't realize that highlighting text will erase the paste buffer, replacing the previous text with what was just highlighted. This still catches me at times, and can be annoying, depending on the complexity of the task I was working on...:-/ But I've done similar stupid things with Ctrl-C/Control-V.
As to intuitive, it took me a good couple days to drill into my dad (A 10 year computing veteran as a user) that Ctrl-C/Ctrl-V would say him time. And I'd get a call months later asking "what's that feature again, that ctrl-key thing". Windows can be pretty fscking annoying, and this is a guy who's used Windows since 3.1. Windows 95, 98, 98SE, Me, and now XP Pro.
I'm no linux apologist, but Windows isn't all you think it's cracked up to be. Personally, I think even the Mac is unintuitive for a new user. Could be I'm just set in my Unix/Windows ways...
I had a look at Elektra, and it's missing what I consider a very vital feature in anything that looks to replace the traditional/etc status-quo: rollbacks and change management/auditing. I need to be able to tell WHO made a change (if you're on a pty as non-root, I can find out who you are), when, and hopefully why. More important, I have to be able to roll that change back when it breaks the system. Preferrably without losing the change, so I can analyze it further, perhaps fixing it. I like the concept of Elektra, and I wish I wasn't getting involved in other OSS projects, because I'd implement this feature for it...
Anyhow, with Novell dropped Yast2 into the OSS fold, I feel the day of decent management tools has come. Yast2 is simply amazing. Combine it with redcarpet/RPMFIND.net, and we're almost there in providing the desktop experience. If only someone could tell me WTF awele-1.0-344.i586.rpm is.:-D
I'll only take exception to your LDAP comment here. Other than the ADSI COM interfaces, Active Directory is nothing more than an LDAP provider. That's the value-add in AD, those com interfaces. But we can agree to disagree.
convert DRIVE:/fs:ntfs
turn off machine
mount bigger drive in machine as IDE0:1.
Boot KNOPPIX from CD.
Open terminal session.
dd if=/dev/hda of=/dev/hdb
KNOPPIX Menu (Utilities I think) QTPARTED.
Resize the filesystem.
Reboot.
For linux, it's the same, except you don't resize the filesystem. If you are using LVM you can add space to a logical volume, and use a filesystem specific tool like resize_reiserfs to resize the filesystem to fill the logical volume without rebooting, without unmounting the filesystem, and CERTAINLY without shutting down your computer (varies by filesystem, of course).
About a month ago, I had to increase the drive size in a Windows2000 HP Omnibook laptop (6.5GB to 40GB), and an XP VMware host on my main server (4GB to 8GB).
I used a slightly modified version of this process, except rather than pull the disk drive, I did this:
smbmount//server/path/mnt/server
dd if=/dev/hda of=/mnt/server/filesystem.img
shutdown.
created new disk (virtual). Installed new drive (laptop).
boot knoppix.
smbmount//server/path/mnt/server
dd if=/mnt/server/filesystem.img of=/dev/hda.
wash, rinse, repeat.
Rather than completely reinstall my work environment from scratch, I simply grew into my new drive, without needing to create new partitions.
When my home directory on my server runs out of disk space, I simply lvextend -L +4GB, and run resize_reiserfs on it. I've got a 250GB drive, half of which is unpartitioned, simply because I have no idea how I'm going to use it all. I could have made it one big partition, but that's chaos, havoc for upgrading. I can wipe out //etc/usr, and I'll still have all my data. My VMware virtual hosts, my databases, my email, my home directory...
Currently you have to use the command line. Given a week, I could probably write a perl-GTK or Java wrapper around all of this and make it a push-button affair. Maybe a month to catch all the fringe cases and take into account scsi devices.
Linux isn't generally better than windows, just some things. It excels at these bare-metal things, unlike windows. Windows has these sorts of products available, but they cost money. PartionMagic and Ghost will never be available commercially for Linux, there's just no need.
You are right. Windows is a great desktop experience. Odds are, however, that if I put you down in front of a freshly installed Knoppix machine, and an XP machine, other than having nearly every application on the web work out of the box perfectly on the XP machine, you'd never know the difference.
Linux has come a long way. Application support is the only thing it really has against it right now. It can compete everywhere, for *MOST* people. Not all, but most.
Other than a uniform installer, and a decent configuration manager, Knoppix does things the Windows-way.
While I like.conf files for their unerring simplicity in the face of database corruption, I would rather have some.XML format that is more easily managed by a uniform configuration manager. Jboss does this with it's containers, and other software is going that route as well. Certainly we can do something uniform across the configuration space that also does configuration management, so that changes can be undone as atomic units. Oh, doing so-and-so broke the machine, restore the configs. (Similar to XP System Restore.).
Microsoft got a lot of things Right with WindowsXP. The OSS world would be well advised to pay attention.
Microsoft WANTED to put NT everywhere, but the DOS and game legacy prevented it. Millenium broke a LOT of things with DOS compatibilty. This was a big wakeup to a lot of vendors that things were coming to a head. But most of the software I ran on Windows95 runs without error, and more stably on XP. I'm really impressed with that. It took them time, too long, in my opinion, but I'm glad that day is here.
Simple. You use wxWindows. Which solves the problem of supporting Windows, MacOS, native X, GTK, and others. There's a difference between picking a Windows toolkit out there from a commercial vendor, and one supported by the open source community. Both Gnome and KDE have more support than Native X does. Neither community is going anywhere, anytime soon. It'd be nice if they could agree to merge at some point, and possibly remain backwards compatible, but that's a pipe-dream.
I'll lay $50 out right now, you can claim it from my grand-children. Gnome and KDE will outlive the X Window System.
We don't need a standard desktop. We need a standard installer and configuration manager more than we need a standard desktop. Much more./etc is almost getting out of control on some of my machines. Almost. No built in version control, dangerous.
The software installation is still a problem, particularly with vendors who don't necessarily keep up to date with packages (Fedora doesn't seem to have this problem, but SuSE does). Yast2 is the most sophisticated configuration and installation tool I have ever used, and coupled with redcarpet package management has left me without needing to deal with too much command-line building, expect some very esoteric projects that I need to work on.
It's not windows. Not yet. And I really hope it never becomes so. When WindowsXP's default is to log the user in as Administrator, and most VIDEO games require administrator access to run (try battlefield:1942 or doom3 sometime), I'm not too keen on storing my valuable data on such a machine.
I have Windows simply to test software I crosscompile on linux, and to run video games, and the rare occasion I need Microsoft Office. But I'm not an ordinary user...
It's all FUD. A Linux fork proper does NOTHING, absolutely NOTHING to affect the fact that the Open Source Solution has finally arrived. Linux is but a kernel. A fork will still be but a kernel. Apache, Samba, Perl, PHP, Compiere, Sql-ledger, sendmail, postfix, postgres, mysql, openoffice, X.org/XFree... These are all pieces of the great Open Source puzzle that enables Linux to prosper... those same projects are leveraged by BSD, Sun JDS, OSX, Windows even to deliver end user value.
Forking linux is a pointless exercise. The Hurd could become fully operational tomorrow and it was have ZERO impact on the open source solution, because it's already here, and growing more rapidly than these vendors can anticipate and react. The OSS desktop can do 99% of what a windows desktop can do today. All it's waiting for is the deluge of applications. And those are coming. Slowly, but surely, they are coming.
These vendors have a right to fear. Postgres is squashing into Oracle and Microsoft territory. SAP-DB and CA's new OSS databases are providing yet another antagonist. Mysql is replacing Access nearly everywhere a small database is needed. The web is firmly entrenched on Apache, and that's only growing (although ASP.net is providing firm new opposition). OpenOffice while still young and immature is firm competition. And Java... well, with the way OSS has been going the past 10 years, and the way the GNU classpath project is going, we'll have an open source Java within 5 years whether Sun likes it or not.
Going with the lowest bidder usually costs you more in the long run than picking someone more expensive, simply because the quality from the more expensive vendor is likely (not always) to be higher in the first place, saving costly remanufacturing or refactoring costs.
Something interesting was related to me yesterday regarding Accounting, and that like computers and high-tech, while there are many CPA's out there, there are few GREAT CPA's out there, and less every year. Attracting those few GREAT CPA's might come down to something simple as having remote access to your Account systems. And maybe instead of dedicated VPN, networking and other infrastructure, you outsource to an ASP. Maybe.
But you're right, it's still nothing new. The only thing the web gives us is a universal, uniform (relatively), ubiquitous interface. There's hope that more applications can move to the thin-client model as the HTML interface matures. And I like that. I like the idea of being able to access my data. Anywhere. But I fear the security implications. I've written elsewhere about OS and download on demand apps on credit cards, so I won't rehash myself, but that's what I'd like to see... for those cases when I feel I can trust the hardware, and want to plug in MY computing environment, rather than use the one provided to me.
I'll second this. I spent YEARS at PTC qualifying specific hardware configurations for support before we finally figured out that we had to target feature sets, not "Compaq Model #1010 w/ AccelGraphics Graphics Engine #2020." It was a PITA, and 99.9% of all functionality worked without error on generic 3d hardware of the day (voodoo, mid, early 1997). Part of that had to do with fewer vendors paying for the certification as the margins in the workstation business collapsed under the weight of the PII flooding the market, but it was a good wake-up call.
Object Design/Excelon/Progress (same company, different names) does it with ObjectStore... It's not a hard problem. If autoconf can figure out how to build bacula on my highly tweaked SuSE 8.2 Linux configuration, it can surely be used to figure out if Random Linux #.1 can support Random App 1.3.
I may not have presented it that way, but that was my end goal. Entire OS, plus data, stored in a credit card form factor. RSA Security already provides credit card two-factor authentication, this is just a small step beyond pendrives.
Dynamic download and local caching of data on your credit card. Secure authentication of transactions via a pin number on the card at all point of sale locations that reads info about your cards off your personal local database.
Easily replaceable. Plug it into a box at home, you can always replace it and reinitialize all your credit card pin numbers with a touch of a button.
Why wouldn't you have to pay for it? I could imagine if they offered some decent fee for bundled usage, office app + accounting app == $25/m + storage costs, but not $100/month for office + $100/m for peachtree accounting... no way. Not viable.
But neither is $0 thin-client computing.
For home owners, once webapps evolve to where something like MSWord is doable, it's just a matter of time before those applications move to the web. Then it's a matter of time before those applications become throw-away. Want to stay on WordWeb10.9, then fine. stay there. There's a whole virtual directory for that. That's doable. People who want to evolve can do so.
Witness the Yahoo email switch. Some people used the new version, others didn't. Can happen the same way with web-apps.
It's finding the right cost/price point that makes it work.
part of the backup problem is that the price/storage ratio on backup media is nowhere near as good as it used to be in the days of 250MB hard drives. When a decent DLT drive plus a dozen $40 tapes are necessary to back up an entry level PC on a decent basis, it never happens. And most end users never backup at all. It's why home raid has become popular. Most users only care about hardware failure, which on board IDE raid1 protects against.
I was talking to a CPA the other day, and yes, real thin-client accounting has yet to take root. I suppose that issues such as security and the dread sword called Sarbanes-Oxley have something to do with that, but time will eliminate most of those barriers, and vendors will come into play delivering those solutions.
Best software, of MAS90/MAS200/Peachtree fame, is already delivering some small business accounting online. A big issue with accounting, however, is availability. A business that deals in transactions is going to be really pissed when their systems are unavailable. Other issues with ASP-style accounting is integration, being able to integrate your manufacturing systems with accounting, inventory control, and sales in a real-time manner.
For small business, you can already get real accounting online. Best Software. For real large manufacturers, it will be some time, if ever.
What I see as potentially working, are thin-clients, DHCP based clients, on a credit card form factor (USB based storage). You jack your credit card into whatever machine you're sitting at, and you access your data. If you need an application, you get it delivered via a mechanism similar to Java Web Start, apps cached locally.
While this doesn't eliminate the trust issue, is the hardware snooping on me, that can be eliminated by ever more powerful palmtop computers, and those sorts of people who care about that, are going to have solutions to that problem. For the vast majority of other people for whom using a library internet terminal isn't a big deal, this is a good thing. Being able to keep your data with you at all times in the event of a network failure is also an attractive selling point (a big problem with the thin-client revolution).
People have been targeting RedHat and SuSE for some time. Other than the abrupt end of Redhat and the emergence of Fedora Core, big deal. Thanks to adherence to the LSB, this isn't as big a deal as it was five years ago.
RPM's built for Redhat 9 typically work without error on my SuSE 8.2 machine, unless it's something designed specifically for system management. This is a cop out.
Oracle is not supported on Fedora Core 2, but I'm running it, thanks to some tricks with tweaking a few/etc files. So what's the big deal?
You are so right, annoying as hell.:-/ I think vendors are a little too tied to checking uname -a and not enough time making sure their libraries are installed properly. Which is where autoconf comes in... maybe it's time installers start running it before committing an install?
* dispatching events
* managing discretely different but cooperating events
Obvious, but yet no so. Reactor/Proactor patterns take care of the first, careful programming is all that can really address the seecond. If I have a select loop, I'll dispatch connecting sockets to a set of worker threads (web server). If I have some log cleanup event that needs to be fired, I could spawn an object with a self-contained thread, from that same Reactor pattern, fire and forget.
Slashdot Mirror
I run k3b along with Evolution, gaim, openoffice, firefox, eclipse, vmware and xmms all on a gnome desktop. They all work well together, look good, are intuitive to use... the lone Piece of Shit I have installed is Acrobat Reader 5, which is a throwback to cheesy 1994 style X-Window interfaces. I'm so looking forward to the 7.0 beta.
So the one commercial application I have installed is also the most mundane, most non-standard widget wise. Funny how that works, huh?
In my 11 years of using X-Windows, middle click has NEVER failed to work with copy/paste.
:-/ But I've done similar stupid things with Ctrl-C/Control-V.
Some people don't realize that highlighting text will erase the paste buffer, replacing the previous text with what was just highlighted. This still catches me at times, and can be annoying, depending on the complexity of the task I was working on...
As to intuitive, it took me a good couple days to drill into my dad (A 10 year computing veteran as a user) that Ctrl-C/Ctrl-V would say him time. And I'd get a call months later asking "what's that feature again, that ctrl-key thing". Windows can be pretty fscking annoying, and this is a guy who's used Windows since 3.1. Windows 95, 98, 98SE, Me, and now XP Pro.
I'm no linux apologist, but Windows isn't all you think it's cracked up to be. Personally, I think even the Mac is unintuitive for a new user. Could be I'm just set in my Unix/Windows ways...
I had a look at Elektra, and it's missing what I consider a very vital feature in anything that looks to replace the traditional /etc status-quo: rollbacks and change management/auditing. I need to be able to tell WHO made a change (if you're on a pty as non-root, I can find out who you are), when, and hopefully why. More important, I have to be able to roll that change back when it breaks the system. Preferrably without losing the change, so I can analyze it further, perhaps fixing it. I like the concept of Elektra, and I wish I wasn't getting involved in other OSS projects, because I'd implement this feature for it...
:-D
Anyhow, with Novell dropped Yast2 into the OSS fold, I feel the day of decent management tools has come. Yast2 is simply amazing. Combine it with redcarpet/RPMFIND.net, and we're almost there in providing the desktop experience. If only someone could tell me WTF awele-1.0-344.i586.rpm is.
I'll only take exception to your LDAP comment here. Other than the ADSI COM interfaces, Active Directory is nothing more than an LDAP provider. That's the value-add in AD, those com interfaces. But we can agree to disagree.
(Sorry, and LDAP user for many many years.)
Assuming you are a windows user:
/fs:ntfs
//server/path /mnt/server //server/path /mnt/server
/etc /usr, and I'll still have all my data. My VMware virtual hosts, my databases, my email, my home directory...
convert DRIVE:
turn off machine
mount bigger drive in machine as IDE0:1.
Boot KNOPPIX from CD.
Open terminal session.
dd if=/dev/hda of=/dev/hdb
KNOPPIX Menu (Utilities I think) QTPARTED.
Resize the filesystem.
Reboot.
For linux, it's the same, except you don't resize the filesystem. If you are using LVM you can add space to a logical volume, and use a filesystem specific tool like resize_reiserfs to resize the filesystem to fill the logical volume without rebooting, without unmounting the filesystem, and CERTAINLY without shutting down your computer (varies by filesystem, of course).
About a month ago, I had to increase the drive size in a Windows2000 HP Omnibook laptop (6.5GB to 40GB), and an XP VMware host on my main server (4GB to 8GB).
I used a slightly modified version of this process, except rather than pull the disk drive, I did this:
smbmount
dd if=/dev/hda of=/mnt/server/filesystem.img
shutdown.
created new disk (virtual). Installed new drive (laptop).
boot knoppix.
smbmount
dd if=/mnt/server/filesystem.img of=/dev/hda.
wash, rinse, repeat.
Rather than completely reinstall my work environment from scratch, I simply grew into my new drive, without needing to create new partitions.
When my home directory on my server runs out of disk space, I simply lvextend -L +4GB, and run resize_reiserfs on it. I've got a 250GB drive, half of which is unpartitioned, simply because I have no idea how I'm going to use it all. I could have made it one big partition, but that's chaos, havoc for upgrading. I can wipe out /
Currently you have to use the command line. Given a week, I could probably write a perl-GTK or Java wrapper around all of this and make it a push-button affair. Maybe a month to catch all the fringe cases and take into account scsi devices.
Linux isn't generally better than windows, just some things. It excels at these bare-metal things, unlike windows. Windows has these sorts of products available, but they cost money. PartionMagic and Ghost will never be available commercially for Linux, there's just no need.
You are right. Windows is a great desktop experience. Odds are, however, that if I put you down in front of a freshly installed Knoppix machine, and an XP machine, other than having nearly every application on the web work out of the box perfectly on the XP machine, you'd never know the difference.
.conf files for their unerring simplicity in the face of database corruption, I would rather have some .XML format that is more easily managed by a uniform configuration manager. Jboss does this with it's containers, and other software is going that route as well. Certainly we can do something uniform across the configuration space that also does configuration management, so that changes can be undone as atomic units. Oh, doing so-and-so broke the machine, restore the configs. (Similar to XP System Restore.).
Linux has come a long way. Application support is the only thing it really has against it right now. It can compete everywhere, for *MOST* people. Not all, but most.
Other than a uniform installer, and a decent configuration manager, Knoppix does things the Windows-way.
While I like
Microsoft got a lot of things Right with WindowsXP. The OSS world would be well advised to pay attention.
This brings up a good question.
Anyone got any recommendations for a tarball installer?
Microsoft WANTED to put NT everywhere, but the DOS and game legacy prevented it. Millenium broke a LOT of things with DOS compatibilty. This was a big wakeup to a lot of vendors that things were coming to a head. But most of the software I ran on Windows95 runs without error, and more stably on XP. I'm really impressed with that. It took them time, too long, in my opinion, but I'm glad that day is here.
Simple. You use wxWindows. Which solves the problem of supporting Windows, MacOS, native X, GTK, and others. There's a difference between picking a Windows toolkit out there from a commercial vendor, and one supported by the open source community. Both Gnome and KDE have more support than Native X does. Neither community is going anywhere, anytime soon. It'd be nice if they could agree to merge at some point, and possibly remain backwards compatible, but that's a pipe-dream.
/etc is almost getting out of control on some of my machines. Almost. No built in version control, dangerous.
I'll lay $50 out right now, you can claim it from my grand-children. Gnome and KDE will outlive the X Window System.
We don't need a standard desktop. We need a standard installer and configuration manager more than we need a standard desktop. Much more.
You've obviously never used Yast2 or anaconda...
:-)
The software installation is still a problem, particularly with vendors who don't necessarily keep up to date with packages (Fedora doesn't seem to have this problem, but SuSE does). Yast2 is the most sophisticated configuration and installation tool I have ever used, and coupled with redcarpet package management has left me without needing to deal with too much command-line building, expect some very esoteric projects that I need to work on.
It's not windows. Not yet. And I really hope it never becomes so. When WindowsXP's default is to log the user in as Administrator, and most VIDEO games require administrator access to run (try battlefield:1942 or doom3 sometime), I'm not too keen on storing my valuable data on such a machine.
I have Windows simply to test software I crosscompile on linux, and to run video games, and the rare occasion I need Microsoft Office. But I'm not an ordinary user...
Good day, sir!
-Chris
I disagree. I was always of the mind that HELPED was equivalent to helped and *helped* was sarcasm.
XML.
If you can't figure out the format of an XML file (with appropriate Schema or DTD) in 15 years, god help you.
It's all FUD. A Linux fork proper does NOTHING, absolutely NOTHING to affect the fact that the Open Source Solution has finally arrived. Linux is but a kernel. A fork will still be but a kernel. Apache, Samba, Perl, PHP, Compiere, Sql-ledger, sendmail, postfix, postgres, mysql, openoffice, X.org/XFree... These are all pieces of the great Open Source puzzle that enables Linux to prosper... those same projects are leveraged by BSD, Sun JDS, OSX, Windows even to deliver end user value.
:-D
Forking linux is a pointless exercise. The Hurd could become fully operational tomorrow and it was have ZERO impact on the open source solution, because it's already here, and growing more rapidly than these vendors can anticipate and react. The OSS desktop can do 99% of what a windows desktop can do today. All it's waiting for is the deluge of applications. And those are coming. Slowly, but surely, they are coming.
These vendors have a right to fear. Postgres is squashing into Oracle and Microsoft territory. SAP-DB and CA's new OSS databases are providing yet another antagonist. Mysql is replacing Access nearly everywhere a small database is needed. The web is firmly entrenched on Apache, and that's only growing (although ASP.net is providing firm new opposition). OpenOffice while still young and immature is firm competition. And Java... well, with the way OSS has been going the past 10 years, and the way the GNU classpath project is going, we'll have an open source Java within 5 years whether Sun likes it or not.
Man, am I an OSS fanboy or what?
Um, lets see here...
e sql-server-7.3...rpm
wget http://ftp.suse.com/pub/suse/9.2/i386/RPMS/postgr
rpm -i postgresql-7.3...rpm
Done.
Going with the lowest bidder usually costs you more in the long run than picking someone more expensive, simply because the quality from the more expensive vendor is likely (not always) to be higher in the first place, saving costly remanufacturing or refactoring costs.
Something interesting was related to me yesterday regarding Accounting, and that like computers and high-tech, while there are many CPA's out there, there are few GREAT CPA's out there, and less every year. Attracting those few GREAT CPA's might come down to something simple as having remote access to your Account systems. And maybe instead of dedicated VPN, networking and other infrastructure, you outsource to an ASP. Maybe.
But you're right, it's still nothing new. The only thing the web gives us is a universal, uniform (relatively), ubiquitous interface. There's hope that more applications can move to the thin-client model as the HTML interface matures. And I like that. I like the idea of being able to access my data. Anywhere. But I fear the security implications. I've written elsewhere about OS and download on demand apps on credit cards, so I won't rehash myself, but that's what I'd like to see... for those cases when I feel I can trust the hardware, and want to plug in MY computing environment, rather than use the one provided to me.
I'll second this. I spent YEARS at PTC qualifying specific hardware configurations for support before we finally figured out that we had to target feature sets, not "Compaq Model #1010 w/ AccelGraphics Graphics Engine #2020." It was a PITA, and 99.9% of all functionality worked without error on generic 3d hardware of the day (voodoo, mid, early 1997). Part of that had to do with fewer vendors paying for the certification as the margins in the workstation business collapsed under the weight of the PII flooding the market, but it was a good wake-up call.
Object Design/Excelon/Progress (same company, different names) does it with ObjectStore... It's not a hard problem. If autoconf can figure out how to build bacula on my highly tweaked SuSE 8.2 Linux configuration, it can surely be used to figure out if Random Linux #.1 can support Random App 1.3.
I may not have presented it that way, but that was my end goal. Entire OS, plus data, stored in a credit card form factor. RSA Security already provides credit card two-factor authentication, this is just a small step beyond pendrives.
Dynamic download and local caching of data on your credit card. Secure authentication of transactions via a pin number on the card at all point of sale locations that reads info about your cards off your personal local database.
Easily replaceable. Plug it into a box at home, you can always replace it and reinitialize all your credit card pin numbers with a touch of a button.
Why wouldn't you have to pay for it? I could imagine if they offered some decent fee for bundled usage, office app + accounting app == $25/m + storage costs, but not $100/month for office + $100/m for peachtree accounting... no way. Not viable.
But neither is $0 thin-client computing.
For home owners, once webapps evolve to where something like MSWord is doable, it's just a matter of time before those applications move to the web. Then it's a matter of time before those applications become throw-away. Want to stay on WordWeb10.9, then fine. stay there. There's a whole virtual directory for that. That's doable. People who want to evolve can do so.
Witness the Yahoo email switch. Some people used the new version, others didn't. Can happen the same way with web-apps.
It's finding the right cost/price point that makes it work.
part of the backup problem is that the price/storage ratio on backup media is nowhere near as good as it used to be in the days of 250MB hard drives. When a decent DLT drive plus a dozen $40 tapes are necessary to back up an entry level PC on a decent basis, it never happens. And most end users never backup at all. It's why home raid has become popular. Most users only care about hardware failure, which on board IDE raid1 protects against.
I was talking to a CPA the other day, and yes, real thin-client accounting has yet to take root. I suppose that issues such as security and the dread sword called Sarbanes-Oxley have something to do with that, but time will eliminate most of those barriers, and vendors will come into play delivering those solutions.
Best software, of MAS90/MAS200/Peachtree fame, is already delivering some small business accounting online. A big issue with accounting, however, is availability. A business that deals in transactions is going to be really pissed when their systems are unavailable. Other issues with ASP-style accounting is integration, being able to integrate your manufacturing systems with accounting, inventory control, and sales in a real-time manner.
For small business, you can already get real accounting online. Best Software. For real large manufacturers, it will be some time, if ever.
What I see as potentially working, are thin-clients, DHCP based clients, on a credit card form factor (USB based storage). You jack your credit card into whatever machine you're sitting at, and you access your data. If you need an application, you get it delivered via a mechanism similar to Java Web Start, apps cached locally.
While this doesn't eliminate the trust issue, is the hardware snooping on me, that can be eliminated by ever more powerful palmtop computers, and those sorts of people who care about that, are going to have solutions to that problem. For the vast majority of other people for whom using a library internet terminal isn't a big deal, this is a good thing. Being able to keep your data with you at all times in the event of a network failure is also an attractive selling point (a big problem with the thin-client revolution).
But if GPL-violations is wrong, it could backfire and they could find themselves on the wrong end of as slander suit.. .
People have been targeting RedHat and SuSE for some time. Other than the abrupt end of Redhat and the emergence of Fedora Core, big deal. Thanks to adherence to the LSB, this isn't as big a deal as it was five years ago.
/etc files. So what's the big deal?
:-/ I think vendors are a little too tied to checking uname -a and not enough time making sure their libraries are installed properly. Which is where autoconf comes in... maybe it's time installers start running it before committing an install?
RPM's built for Redhat 9 typically work without error on my SuSE 8.2 machine, unless it's something designed specifically for system management. This is a cop out.
Oracle is not supported on Fedora Core 2, but I'm running it, thanks to some tricks with tweaking a few
You are so right, annoying as hell.
Threads are useful for only two things:
* dispatching events
* managing discretely different but cooperating events
Obvious, but yet no so. Reactor/Proactor patterns take care of the first, careful programming is all that can really address the seecond. If I have a select loop, I'll dispatch connecting sockets to a set of worker threads (web server). If I have some log cleanup event that needs to be fired, I could spawn an object with a self-contained thread, from that same Reactor pattern, fire and forget.