Re: [TUHS] Technical decisions based on political considerations [was Re: AT&T Research]

Amen to both comments. Especially Intel, I'm so happy to not be dealing with them. Just an example, they used Netapp's mirroring stuff - stuff that says in big bold letters "Mirrors are read only, you may not write to them". Of course they wrote to them and it didn't go well. They were using BitKeeper and they cloned a repo to a mirror. A clone in BK is basically ( cd from; tar cf - . ) | (mkdir to; cd to; tar xf -; bk repocheck) It is more complicated than that, it doesn't do a tar, it finds all the SCCS files and transfers those and a handful of etc files. And it has a bill of materials file that lists all the SCCS files. So a repocheck is sort of like find . -name 's.*' | check that list against the bill of materials When we unpacked the files and went to go look for them, half the files that we just wrote were "not there". They were there but there were no entries for them in the directory so it looked like they were not there. Intel said that BitKeeper was broken. For 3 months. After I gave them scripts that replicated what BitKeeper was doing but had no BitKeeper in them. After tuning those scripts so their so-called filer validation team could use them as a test system to verify that the filers worked. As a kernel guy, I did the most in depth work in file systems. I know what I'm talking about. But Intel said it was all Bitkeeper's fault. It wasn't, it was just that BitKeeper was the only application they had that did integrity checks. They finally backed down when I called Steve Kleinman who was CTO at Netapp and my mentor at Sun, he immediately said yeah, I know, it's us, that mirror shit sucks. And he called Intel. Intel is just an awful company. I know they pay Clem's bills, go Cleam, but Intel sucks. On Thu, Jul 23, 2020 at 12:02:04AM -0600, arnold(a)skeeve.com wrote: -- --- Larry McVoy lm at mcvoy.com http://www.mcvoy.com/lm