Neobuntu
December 17th, 2007, 09:38 PM
MODERATORS: Please put this where it may be considered by the main decision making developers. Thank you.
The best test for a broad based public distribution is the actual time it saves us getting the job(s) done. Using it (first and foremost) and second to that, installer enhancement. Thankfully installing is fast and easy, unlike Windows.
When it comes to inter-release upgrading (not dist-upgrading), this is very important and indeed lets you stick with your install nearly indefinitely. Still, it's current and up to date. YOU decide when to change. Ubuntu; with is large community, excels at this. Even when (especially when) the poop hits the fan. Like a wide-spread bug. It is fixed and fixed FAST!
Now dist-upgrading to a new release is a different matter. If you are running the Debian "stable" tier as a server, then maybe. Go ahead then and dist-upgrade to the new "stable". Yet, with others such as Ubuntu/Kubuntu, NO WAY! Clean install it. Let me explain.
Now, great effort is done with Ubuntu to enhance and "manage" the dist-upgrade to the next "release". Yet just because some have had an easy, managed (or not) dist-upgrade, does NOT prove it is better. I know, you are like me and want to save all your customizing. It's hard to remember what you did and we certainly do not want to waste time doing it all again. Thus, the dream of a managed Ubuntu dist-upgrade. Please keep reading and see below about enhancing a clean install, instead.
So, why is managed/enhanced dist-upgrading not better? First, it's a gamble, especially within the first period of time; after a new release. By the time the managed dist-upgrade is stable, the next release has moved on (with the community by in large). Old stuff is not what Ubuntu (and family) is all about. That's Debian stable. If you are one of the many with a problem dist-upgrading into the next release, you'll have wasted your time falling back to a clean install (and restoring your stuff) anyway. So already, backing up and clean installing is faster than upgrading. Restore that which you can live without is fast and straight forward too. So then, you may have to clean install anyway. It's your choice but there's more.
Also, when you have a bug, if you are not starting from the current release and as a clean install, you can't know if that has contributed to your problem. If you clean install, it's one less thing to worry about. You don't have to wonder if it is your one and only old install COMBINED with a managed-dist-upgrade that is causing a bug/failure. Can you see how starting with EVERY users tweaked system differences is a recipe for disaster with support? Also, you are then up with the developers/community and what they are running. When major fixes come through, you're good to go. It is safety in numbers. Large numbers.
Often the new installer does things that you (or the devs) might miss, with a dist-upgrade. Things change-over to new methods, features and benefits. Often the clean installer is the least time wasted with these changes. Especially with foundational matters.
Mostly, we CAN'T debug what we haven't created yet! Since only the installer is static, and the distro is actually in flux (or dynamic) with each (inter release) upgrade, a managed dist-upgrade CAN NOT fully account for a proper conversion until it is adjusted. What has the distro then become. This is why too many users can not up their release; just after the new one is deemed "released". Even a 75% initial success rate here, is a failure. This drives real people away. A 99% success when it's old, does not help the real problem. It wastes effort (when releases are new).
You may have to read that one again.
Contrastingly, a clean install can accept the newest (inter) upgrades immediately and stability is the norm. This also makes it easier for those upgrades to work (if everyone starts with a clean install). It seems like less effort on both sides, to me.
Now perhaps we should further enhance/manage the clean install instead of the dist-upgrade. Eh?
The first help that comes to mind, is how we can utilize our old home directory. Although, there are many ways to do this. Maybe a simple backup option(of your old stuff) would be welcome in the clean installer. Then a simple restore perhaps. options abounding. Even this can lend poorly toward a release upgrade way of thinking. It is probably best, to better think of these things as a managed, clean install.
How much vain effort is wasted toward dist-upgrading when perhaps clean installing is what can be optimized? Said again in another way, can dist-upgrading Ubuntu now be stabilized (from driving away newbies) before it is too old in the release cycle? It may not be realistic for a Six month release cycle. It may be wise only when your release is woefully out of date. Like by almost two releases (with Ubuntu), so that you'd still be behind by almost one release else you'd risk instability dist-upgrading).
It occurs to me also, this would give some freedom from legacy-change issues, to the developers as well. AKA, faster development cycles with better stability. Isn't that the great challenge?
Take a page from Debian. If you want to test and play with unstable apps, one starts by better installing "stable" because it's installer is going to give you (and most) little trouble (stable). Then the technically inclined then dist-upgrade into possible major instability for "testing" and "unstable".
As Ubuntu stabilizes a newer Debian and makes it MORE user friendly, it is the stable and time saving side we should base the upgrading upon. Not the testing, push the envelope side. This idea of "Oh, it wasn't supposed to have much stability" is not winning any friends to open software IMHO. We have Sid for that.
Rule number one: Know who (or what) you are.
The best test for a broad based public distribution is the actual time it saves us getting the job(s) done. Using it (first and foremost) and second to that, installer enhancement. Thankfully installing is fast and easy, unlike Windows.
When it comes to inter-release upgrading (not dist-upgrading), this is very important and indeed lets you stick with your install nearly indefinitely. Still, it's current and up to date. YOU decide when to change. Ubuntu; with is large community, excels at this. Even when (especially when) the poop hits the fan. Like a wide-spread bug. It is fixed and fixed FAST!
Now dist-upgrading to a new release is a different matter. If you are running the Debian "stable" tier as a server, then maybe. Go ahead then and dist-upgrade to the new "stable". Yet, with others such as Ubuntu/Kubuntu, NO WAY! Clean install it. Let me explain.
Now, great effort is done with Ubuntu to enhance and "manage" the dist-upgrade to the next "release". Yet just because some have had an easy, managed (or not) dist-upgrade, does NOT prove it is better. I know, you are like me and want to save all your customizing. It's hard to remember what you did and we certainly do not want to waste time doing it all again. Thus, the dream of a managed Ubuntu dist-upgrade. Please keep reading and see below about enhancing a clean install, instead.
So, why is managed/enhanced dist-upgrading not better? First, it's a gamble, especially within the first period of time; after a new release. By the time the managed dist-upgrade is stable, the next release has moved on (with the community by in large). Old stuff is not what Ubuntu (and family) is all about. That's Debian stable. If you are one of the many with a problem dist-upgrading into the next release, you'll have wasted your time falling back to a clean install (and restoring your stuff) anyway. So already, backing up and clean installing is faster than upgrading. Restore that which you can live without is fast and straight forward too. So then, you may have to clean install anyway. It's your choice but there's more.
Also, when you have a bug, if you are not starting from the current release and as a clean install, you can't know if that has contributed to your problem. If you clean install, it's one less thing to worry about. You don't have to wonder if it is your one and only old install COMBINED with a managed-dist-upgrade that is causing a bug/failure. Can you see how starting with EVERY users tweaked system differences is a recipe for disaster with support? Also, you are then up with the developers/community and what they are running. When major fixes come through, you're good to go. It is safety in numbers. Large numbers.
Often the new installer does things that you (or the devs) might miss, with a dist-upgrade. Things change-over to new methods, features and benefits. Often the clean installer is the least time wasted with these changes. Especially with foundational matters.
Mostly, we CAN'T debug what we haven't created yet! Since only the installer is static, and the distro is actually in flux (or dynamic) with each (inter release) upgrade, a managed dist-upgrade CAN NOT fully account for a proper conversion until it is adjusted. What has the distro then become. This is why too many users can not up their release; just after the new one is deemed "released". Even a 75% initial success rate here, is a failure. This drives real people away. A 99% success when it's old, does not help the real problem. It wastes effort (when releases are new).
You may have to read that one again.
Contrastingly, a clean install can accept the newest (inter) upgrades immediately and stability is the norm. This also makes it easier for those upgrades to work (if everyone starts with a clean install). It seems like less effort on both sides, to me.
Now perhaps we should further enhance/manage the clean install instead of the dist-upgrade. Eh?
The first help that comes to mind, is how we can utilize our old home directory. Although, there are many ways to do this. Maybe a simple backup option(of your old stuff) would be welcome in the clean installer. Then a simple restore perhaps. options abounding. Even this can lend poorly toward a release upgrade way of thinking. It is probably best, to better think of these things as a managed, clean install.
How much vain effort is wasted toward dist-upgrading when perhaps clean installing is what can be optimized? Said again in another way, can dist-upgrading Ubuntu now be stabilized (from driving away newbies) before it is too old in the release cycle? It may not be realistic for a Six month release cycle. It may be wise only when your release is woefully out of date. Like by almost two releases (with Ubuntu), so that you'd still be behind by almost one release else you'd risk instability dist-upgrading).
It occurs to me also, this would give some freedom from legacy-change issues, to the developers as well. AKA, faster development cycles with better stability. Isn't that the great challenge?
Take a page from Debian. If you want to test and play with unstable apps, one starts by better installing "stable" because it's installer is going to give you (and most) little trouble (stable). Then the technically inclined then dist-upgrade into possible major instability for "testing" and "unstable".
As Ubuntu stabilizes a newer Debian and makes it MORE user friendly, it is the stable and time saving side we should base the upgrading upon. Not the testing, push the envelope side. This idea of "Oh, it wasn't supposed to have much stability" is not winning any friends to open software IMHO. We have Sid for that.
Rule number one: Know who (or what) you are.