Re: [PATCH 2/2] block: create ioctl to discard-or-zeroout a range of blocks

On Tue, Mar 15, 2016 at 4:52 PM, Dave Chinner [off-list ref] wrote:

It is pretty clear that the onus is on the patch submitter to
provide justification for inclusion, not for the reviewer/Maintainer
to have to prove that the solution is unworkable.

I agree, but quite frankly, performance is a good justification.

So if Ted can give performance numbers, that's justification enough.
We've certainly taken changes with less.

And with your "we should _not_ do this" argument, the onus is clearly on you.

"Google uses this" is not sufficient justification.

Not per se, no, but it's a very traditional and time-honored model for
"should we merge this".

It's traditionally been things like "Redhat merged it in their distro
kernel, because they have customers that want/need it". That is *the*
reason many big projects got merged, ranging from filesystems to
drivers etc. Take reiserfs, for example: it got merged because SuSE
was actively using it.

So "this feature is being used in real life" is a big hint that the
standard upstream kernel may be missing something important. People
arguing against things like that has been a big problem in the past.
It took people _years_ to get over the whole Android thing. We need to
merge stuff that people are using and depend on, because _not_ merging
them just causes more and more distance between peoples kernels, and
makes it even harder to merge in the future.

I do agree that we want to have hard numbers.

And I do think that we should strive for the whole "we want to merge"
phase to be a time when we also look at "can we improve the
interfaces". But that can - and often does - go too far. Again, we had
_years_ of pointless masturbation over the whole Android thing, just
because people were making up new interfaces.

               Linus