One of the problems with recommending Ubuntu to potential future users, especially those not particularly given to technical endeavours, is that there is a chance that upgrades will break their machine, and they'll have to pay or otherwise coerce some knowledgeable person into fixing them.
In my limited experience of running successive versions of Ubuntu since 8-something on a couple of different laptops, this chance is quite high. I'm not sure if I'm just unlucky with the hardware that I'm using, or if it's a result of the higher-than-average number of packages I have installed, or if upgrades are just typically problematic.
So I'd like to know the likelihood, for a casual user, of doing a release upgrade, for example from 10.04 to 10.10, without experiencing any regression bugs.
Obviously this is dependent on the hardware that people are running. Canonical seems to be making some efforts towards collecting data on this, for example with the "I am affected by this bug" checkbox on their issue tracker, and with the laptop compatibility reports, but I've not seen anything comprehensive.
I'm hoping for an objective reference here, for example a study carried out by relatively unbiased individuals. However, anecdotal evidence is probably useful too.
I'd love for someone to come up with some more numbers on this, here's my little analysis:
There are around 100 bugs tagged "regression-release", that are of "High" importance. A further three are marked "critical", two of those affect a bunch of specific ThinkPad models.
Have a look at the regression tracker, note only the ones that are tagged "regression-release".
There are quite a lot of them, of course, but then the average bug affects only a few users (now that's a number I'd like to know). Note that pretty much all of the 'serious' bugs have to do with some specific piece of hardware, and thus wouldn't affect "well supported" hardware platforms.
Take the time to sort this list by importance and read some of the descriptions. To me at least, many of them appear very minor, or affect only a small set of users. But I can't be the judge of that.
The likelihood of a normal user experiencing a release upgrade regression is almost impossible to estimate. The hardware platforms vary immensely.
In my experience, everything works perfectly all the time. See how worthless that information is. ;-)
I fear that's all I've got. It's really not a problem that is discussed much, which is why I have to doubt it exists at all. People usually only investigate such things in detail, create comprehensive statistical analysis, if many users are affected; due to the way Canonical and the community test before they release, this appears to be rare.
We do see a lot of discussion on this topic everytime a new release has come out. Most experienced users prefer a fresh installation to an upgrade, and indeed there always are some proplems from incomplete updates or from regressions that may not have come up when you had done a new installation. Still, this recommendations are not based on solid statistical analyses that would be needed to really tell.
Stefano Palazzo has already pointed out why it is difficult, or maybe impossible to perform such an analysis. Too many individual factors had to be considered as no system is set up like the other. We can only have a look at what people post in the support forums to get a vague idea what problems may arise.
No problems on upgrade
On a not too much individually adapted system where software was only installed from Ubuntu repositories and no proprietary drivers are needed there are only very few (if any) reports on defects after an upgrade. This leads me to believe that an upgrade of such a system is nearly 100% safe. Personally I never had any problems when upgrading such a system.
Upgrade fails
When proprietary hardware drivers or packages from sources other than the Ubuntu repositories were installed still, in most cases even this is very cleanly handled by an upgrade. However sometimes then an upgrade may cause problems where additional configuration/repair may be needed. We also have to consider those cases where both, an upgrade failed, and a fresh install after that did not work either because of unsupported hardware.
Subjective factors from supporters or experienced users come in because they are more likely to have highly adapted and individually configured systems therefore being much more affected by adverse effects from an upgrade. This explains why for them it may be less time consuming to simply perform a fresh install.
What to recommend?
Ubuntu allows to keep most of your individual settings even if you performed a fresh install. Also the time needed for a fresh install is in the range of 20 min. as compared to up to 2 hrs for an upgrade. Therefore a fresh install seems very attractive indeed.
This is especially the case for basic installations with only few additional packages where we don't have to do not much additional work. But it is exactly these systems that also would upgrade smoothly.
On the other hand, if your system is very individually configured with many additional packages loaded then you would probably save much time when upgrading. If you are unlucky and the upgrade failed you still haven't lost the option to do a fresh install but you may loose a lot of the saved time to find out if your system can be repaired or not.
From a practical view, recommend an upgrade when:
Recommend a fresh installation when:
This is entirely my personal view biased by my opinion that it does not really matter much what you do.
It's an interesting question, but also very hard to answer, among other things since "trouble free" is a quite subjective metric.
But we can try to triangulate some numbers:
The likelihood of having an upgrade (any upgrade) work perfectly for everyone is zero.
This is independent of which operating system we are talking about, it is merely a function of the enormous number of combinations of people, skills, use cases, hardware and software that exists in the world.
See for example
Microsoft employs 9,000 testers who test daily builds across thousands of hardware and software combinations for years before a major relase; probably noone else in the industry puts as much resources into testing as Microsoft does.
That doesn't stop problem upgrade "Windows 7" from returning 300 million Google hits (the numbers vary, that's what I'm getting at my location at the moment).
Apple presumably has a simpler job because they tightly control the very small number of hardware combinations their OS is intended to run on. Still, problem upgrade "OS X" is worth 8 million Google hits.
Ubuntu (and Linux in general) faces the same problems as Microsoft by targeting close to every hardware combination under the Sun; but unlike Microsoft they are doing so with a fraction of the resources of the major players. problem upgrade Ubuntu gives 21 million hits, while problem upgrade Ubuntu 10.10 is ~3 million.
These are obviously highly unscientific, rough and ready numbers (your Google numbers will probably be different - it varies with location, the exact phrasing of the search, and quite possibly the phase of the moon...), but I think the relative frequency of complaints can still be a better indicator than random guessing.
We need to weigh complaint frequencies against usage frequencies, hard numbers are again hard to come by.
But a random source close at hand says that Win 7 is used by 24%, OS X by 8% and Linux by 1.5% of some unspecified population. (I have no idea what Ubuntu's market share is of the Linux total, but it's almost certainly less than 100% :)
Combining these numbers with my own subjective experience and a vast quantity of anecdotes, hearsay and urban legends, I'm comfortable in believing that
Personally I always wait a few months after a release before I do an upgrade (whether it's Windows, Mac or Ubuntu) so the worst bugs can be ironed out, and I mostly have a reasonable trouble-free upgrade. And it seems to be getting better at every release, at least on the kind of hardware I am using.
To future users I would suggest that unlike romantic partners, nobody will really mind if you flirt with several operating systems at the same time.
So as long as you can spare some room on your primary drive, you can install Ubuntu as dual boot and try out an extramarital OS affair at very little risk.