Obviously there are no features that are required to display a web page that are missing on 10.5. Browsers were developed and ran perfectly fine on there. Some newer features might not work, or some newer optimizations. But remember, keeping around all the old code paths, that check for particular features, can lead to more development cost, more testing cost, more bugs, more security vulnerabilities.
Another example is security features. Snow Leopard (10.6) introduced some new security features, like stack protection and more sandboxing. Chrome is heavily focused on security, and they may not be able to provide the full isolation that they would like on 10.5.
I develop software that should by all rights work fine on 10.5. And nothing that we do can't be made to worth there. But it takes more work; I develop on 10.7, we test heavily on 10.6 and 10.8, and when we go to test on 10.5, we find things that are broken not because they're impossible, but just because some interface is different so we need to add a new code path to handle that case. All that extra testing and development is expensive, and there are increasingly few people still on 10.5, so the benefit is continually shrinking.
And being already set up for testing does not make it free. Hardware dies, and you have to replace it. More and more developers move on to newer OSs, and it becomes harder for them to get all of their development environment working in the older OS to find and fix bugs.
Have you ever tried developing a product with backwards compatibility across 4 different OS releases? It is by no means easy, it limits what you can do, and it makes your software more complicated and bug prone.
Don't think of it as having to make a conscious decision to stop. It's more that you have to make a conscious effort, that gets greater as time goes on, to continue supporting the older release. More libraries that you use start dropping compatibility. More new features have to be ifdef'd out, leading to divergence in what your software supports on different systems. More effort has to be put into keeping the build system running. And more manual testing needs to be done.
(1) Being "already set up for testing" for an older OS version
and
(2) continuing to use the testing setup for that version in addition to ever increasing new test setups (for 10.6, 10.7. 10.8 etc) and codepaths
is not the same thing.
As you get newer OSs to support, maintaining older testing setups means more work, and more IFDEFS and such in the code.
Putting more care in the LATEST offering. Putting more care in older versions if a loosing proposition.
And it's not just a "throw more people / money at it" problem. After some point the complexity of legacy support grows exponentially and the code suffers (or gets hold back).
>Car makers don't stop making spare parts a couple of years after a new model comes out. If they did no one would buy them.
5 years in computing is like 20 car years...
Here's a discussion from some FireFox developers about it. Sounds like they want to drop support for 10.5 as well:
https://groups.google.com/forum/?fromgroups=#!topic/mozilla....
Uniques over this period were between 2-3 million.
I'm part of that small percentage although I'm trying to change to Safari!
But it's an even smaller slice when looking at who's using 10.5. Fraction of a fraction of a fraction of users (Mac > Chrome > 10.5).
The total percentage of Mac+iOS across the total uniques was quite high at around 25% and like your figures, a huge portion of this was from iOS users, not Mac users.
Never thought to do a run down by OSX version but I agree that the numbers would be very small.
And FWIW, Apple requires 10.7.4 for Safari.
Update: in a by no means statistically valid way, I pulled the stats from Google Analytics for my website over the last 30 days... 186,883 Mac visits (if it shows where things are going, iOS had 814,201 visits!), with 33,934 using Intel 10.5.X. This is higher than I was predicting (18.16% of total Mac visits). It's actually about 3 times higher than the number of people on 10.8. However, most of these people are using Safari and just 2,040 were Chrome users( 1.09% of total Mac visits).
If anyone with a more Mac heavy site has numbers, I'd be curious.