Date: Sun, 16 Dec 2018 18:15:33 -0700 From: Warner Losh <imp@bsdimp.com> To: "Bjoern A. Zeeb" <bzeeb-lists@lists.zabbadoz.net> Cc: "Rodney W. Grimes" <freebsd-rwg@pdx.rh.cn85.dnsmgr.net>, freebsd- <arch@freebsd.org>, George Neville-Neil <gnn@neville-neil.com> Subject: Re: A proposal for code removal prior to FreeBSD 13 Message-ID: <CANCZdfojkGv-xYR%2BzvGfVcrmTyB0BnC61p65giooGVAhtyJCaQ@mail.gmail.com> In-Reply-To: <261A6437-3ECC-43FF-ADA2-EE430477BB92@lists.zabbadoz.net> References: <201812161645.wBGGj7qn092076@pdx.rh.CN85.dnsmgr.net> <261A6437-3ECC-43FF-ADA2-EE430477BB92@lists.zabbadoz.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Dec 16, 2018 at 1:45 PM Bjoern A. Zeeb < bzeeb-lists@lists.zabbadoz.net> wrote: > I=E2=80=99ll ignore the =E2=80=9Cturning FreeBSD into change-management > procedures=E2=80=9D part. > OK. I can't let that slide by without comment. What we've had for the past 10 years hasn't worked. Evidence: there's a lot of junk in the tree. People have been afraid to deprecate because there's too much friction. Why's that? Because when people just got a bug up their butt and deleted stuff, people complained (and rightly so). So we, as a project, have been afraid to do anything. I started breaking that log-jam about 6 or 9 months ago when I started the ball rolling on getting rid of the really old arm bits. I learned a lot from that. We learned a lot from the 10/100 stuff. I've learned from other things big and small. That's why I wrote up the evaluation criteria: to try to move to a data-driven process. Sadly, we have no telemetry in our systems due to longly held notions that it's too invasive to do it, so we didn't have it, even on an opt-in basis. We have very little usage data. Absent, that we have to guess. Every single set of guesses I made in my deprecation quest has had at least one flaw in it. These flaws, however, have often been brought to light by socializing the plan widely and letting people comment. In some cases, the comments were addressed as "you don't understand, let me explain it" and in other cases it was "I didn't understand, let me amend my proposal". In both cases the proposal was better after the review. But it's tricky... without the proper guidance, these discussions can go off into the weeds quickly (so people have been reluctant to do them). The key, imho, is making them of small enough scope that all the issues that arise can be dealt with, yet big enough to make it worth the other costs. Honestly, it's gone much more smoothly than I feared when I started. The more I've talked to people (not just developers, or the echo chamber of people I talk to a lot), the more I find the weird places FreeBSD is in use, the more data I gather, the more that I can understand what's going on. And the more I've talked to people about what I'd like to do, the more they are understanding when things don't go like they'd hoped because they know they have had their say fairly considered. So what I've started writing up isn't some mandatory straight jacket for the project, but instead a set of best practices. This has worked well in the past, that hasn't worked, etc. Everybody wants to do their job, including deprecation, with a minimum of friction. But that doesn't necessarily mean "just do it": that path just back loads the friction to after the commit and leaves too much hard feelings. And we don't want to talk it to death before the commit: nobody wants that either. And we've been stuck, as a project, for a long time. I'm hoping to break the log jam and find some guidelines that will help us, as a project talk to all the stakeholders about the issues in a productive way that lets us move forward= . So this isn't at all about process for process sake, nor about creating a new straight jacket to replace the old one of fear that we've worn for too long. It's about using an open approach to hopefully solve this problem once and for all so we can remove what we need to in a healthy and functional way. Sorry for the bit of a rant, but I've been working on this now for a while and don't want to see us return to the bad-old-days. Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfojkGv-xYR%2BzvGfVcrmTyB0BnC61p65giooGVAhtyJCaQ>