The LSB and its Current Implementation

Ulrich has some problems with the LSB standard. He makes some outstanding points and is clearly an extremely smart guy. The problem as I see it is that the current LSB is broken and badly broken at that. When the answer to a bug report is “use a slow uni-processor machine, it is known to work there”, someone should be fired. A certification with a date of 15-Sep-2005 should not be done on a 300MHz Pentium as it's not indicative of what people are running. Is all lost though? Should the LSB just be completely dropped? Maybe this particular group/spec should be, since they are clearly not doing a good job, but it should be clear on why companies like Oracle want to ship a single binary that works everywhere. If we're serious about mainstream Linux adoption, this is something that I think has to happen. Requiring companies to build for each distro is dumb. It means multiple build environments, additional testing, additional resources. In the end, it means that either the company won't ship on Linux or the company will only support a specific distribution or two. That's bad for everyone. I don't have a solid answer (and from the looks of it, no one does) but this is a problem that needs to be solved.
–jeremy
LSB, Linux