It's the testing, not the package building

It's the testing, not the package building

> Look at SAP, for example. They are certainly not clueless wrt linux, and they make separate releases for errata kernels. I bet they have reasons to do so.

I have no idea why SAP have chosen this way of doing things, but there are plenty others that didn't: Oracle, NVidia, Cisco, Macromedia - they all ship just one version of their proprietary products that work across MOST Red Hat flavours, including weird things like latest development Fedoras. So, maybe SAP aren't as good as they'd like to think they are.

I understand where you're coming from with your arguments, just like many other posters. It does come down to being competent as an ISV and not doing things that you aren't supposed to be doing, especially in the kernel space (where most breakage appears to happen). But, I'll repeat my main point again: the author of the article greatly exaggerates the difficulty of doing this. The number of products and RPMS alone doesn't necessarily mean much. I agree that supporting means doing more than just building RPMS, but there is always a few (very similar) flavours that have close to 90% of the market - the rest is just noise from proprietary vendors' point of view. Whether that leaves me or you without that proprietary software is something they usually don't concern themselves with, if the numbers are small enough.