On 12/15/20 6:12 PM, Joshua Kramer wrote:>> I don't think there will be a course change either, but for different >> reasons. The motivation isn't "cashing/selling out". It's... actually the >> stated motivation >> https://www.redhat.com/en/blog/faq-centos-stream-updates#Q2 > > First, I will note that I think the idea of creating *a version of* > CentOS that is called "Stream", with the intent that it leads RHEL by > a bit, is a GREAT idea, for exactly the stated reasons! > > There's one problem I have with this asserted motivation. Stream is > not being done as *a version of* CentOS. It is being done as *THE* > CentOS, which means you're discontinuing point releases. As far as > "maintaining CentOS point releases to follow RHEL"- this is what is > being discontinued. How much money, in developer time and other > incidentals, does this cost RedHat per year? Of course this is a > proprietary number. But let's imagine that this number is $250k per > year. Out of what was it, about $433M of profit (2019)? So it would > cost RedHat 0.06% of profit to hire more developers to keep issuing > CentOS point releases. > > What does RedHat "buy" in return for spending 0.06% of its profit on > maintaining point releases? > -Community trust and goodwill. Those members of the community that > cannot afford RedHat licenses for whatever reason still know that the > #1 player in the Linux marketplace still has their back. Then when > those folks move on to enterprises that can afford RH licensing (and > in some cases demand it), will select RedHat because of this trust and > goodwill. They will be highly likely to recommend other RedHat > products- since it all "works together" and they'll know RHEL (i.e. > CentOS) well. Also note that this trust and goodwill means > "convenience", even within enterprises that have a large budget with > RedHat. If I have a project and I want to spin up 100 OS instances > just for the heck of it, I can. I don't need to ask anyone, I don't > need to reserve or download any entitlement key files. I don't need > to debug weird problems when entitlement key files don't work. > -Control of part of the ecosystem. Those companies that build their > products to run on RHEL (or in RHEL containers) are able to (and > encouraged to) certify those products on RHEL because they are able to > use CentOS. > > But more to the point, what does RedHat LOSE by saving 0.06% of its > profit? The damage to community trust and goodwill far exceeds the > gains that would be realized if the status quo were kept in place. > Yes, it's true that many of the folks who used CentOS would never turn > into paying customers. But due to this situation, you have thousands > of system administrators who are actively looking to completely > abandon the RedHat ecosystem altogether. When it comes time to > recommend products... they aren't going to recommend RHEL. They > aren't going to recommend JBoss, or Fuse, or 3Scale API management. > It's clear that RedHat can't be trusted with some parts of its > portfolio, so why should we trust ANY of its products?So don't trust them. Move to something else if you think something is better.> > If it is 100% factually correct that the ONLY motivation for killing > point releases is the stated motivation, then it's just a simple > matter of finding a spare $250k (or whatever that cost is) from the > almost-half-a-billion dollar corporate coin purse. The return on > investment has been, and will continue to be, immeasurable...$250K is not even close. That is one employee, when you also take into account unemployment insurance, HR, medical insurance etc. now multiply that by 8. Now, outfit those 8 employees to work from home .. all over the world, different countries, different laws. .. THEN buy 30 machines minimum (servers, not workstations) for building and testing, buy a service contract for those 30 machines, host the bandwidth required to sync out to 600 worldwide servers. We need all the CI machines .. that is a bunch of blade servers for that. They need service contacts too. In any event it doesn't matter. The decision is made. If people don't want to use CentOS Stream, then don't. The decision is not changing.
On Tue, Dec 15, 2020 at 7:41 PM Johnny Hughes <johnny at centos.org> wrote:> On 12/15/20 6:12 PM, Joshua Kramer wrote: > >> I don't think there will be a course change either, but for different > >> reasons. The motivation isn't "cashing/selling out". It's... actually > the > >> stated motivation > >> https://www.redhat.com/en/blog/faq-centos-stream-updates#Q2 > > > > First, I will note that I think the idea of creating *a version of* > > CentOS that is called "Stream", with the intent that it leads RHEL by > > a bit, is a GREAT idea, for exactly the stated reasons! > > > > There's one problem I have with this asserted motivation. Stream is > > not being done as *a version of* CentOS. It is being done as *THE* > > CentOS, which means you're discontinuing point releases. As far as > > "maintaining CentOS point releases to follow RHEL"- this is what is > > being discontinued. How much money, in developer time and other > > incidentals, does this cost RedHat per year? Of course this is a > > proprietary number. But let's imagine that this number is $250k per > > year. Out of what was it, about $433M of profit (2019)? So it would > > cost RedHat 0.06% of profit to hire more developers to keep issuing > > CentOS point releases. > > > > What does RedHat "buy" in return for spending 0.06% of its profit on > > maintaining point releases? > > -Community trust and goodwill. Those members of the community that > > cannot afford RedHat licenses for whatever reason still know that the > > #1 player in the Linux marketplace still has their back. Then when > > those folks move on to enterprises that can afford RH licensing (and > > in some cases demand it), will select RedHat because of this trust and > > goodwill. They will be highly likely to recommend other RedHat > > products- since it all "works together" and they'll know RHEL (i.e. > > CentOS) well. Also note that this trust and goodwill means > > "convenience", even within enterprises that have a large budget with > > RedHat. If I have a project and I want to spin up 100 OS instances > > just for the heck of it, I can. I don't need to ask anyone, I don't > > need to reserve or download any entitlement key files. I don't need > > to debug weird problems when entitlement key files don't work. > > -Control of part of the ecosystem. Those companies that build their > > products to run on RHEL (or in RHEL containers) are able to (and > > encouraged to) certify those products on RHEL because they are able to > > use CentOS. > > > > But more to the point, what does RedHat LOSE by saving 0.06% of its > > profit? The damage to community trust and goodwill far exceeds the > > gains that would be realized if the status quo were kept in place. > > Yes, it's true that many of the folks who used CentOS would never turn > > into paying customers. But due to this situation, you have thousands > > of system administrators who are actively looking to completely > > abandon the RedHat ecosystem altogether. When it comes time to > > recommend products... they aren't going to recommend RHEL. They > > aren't going to recommend JBoss, or Fuse, or 3Scale API management. > > It's clear that RedHat can't be trusted with some parts of its > > portfolio, so why should we trust ANY of its products? > > So don't trust them. Move to something else if you think something is > better. > > > > > If it is 100% factually correct that the ONLY motivation for killing > > point releases is the stated motivation, then it's just a simple > > matter of finding a spare $250k (or whatever that cost is) from the > > almost-half-a-billion dollar corporate coin purse. The return on > > investment has been, and will continue to be, immeasurable... > $250K is not even close. That is one employee, when you also take into > account unemployment insurance, HR, medical insurance etc. now multiply > that by 8. Now, outfit those 8 employees to work from home .. all over > the world, different countries, different laws. > > .. THEN buy 30 machines minimum (servers, not workstations) for > building and testing, buy a service contract for those 30 machines, host > the bandwidth required to sync out to 600 worldwide servers. > > We need all the CI machines .. that is a bunch of blade servers for > that. They need service contacts too. > > In any event it doesn't matter. The decision is made. If people don't > want to use CentOS Stream, then don't. The decision is not changing. > _______________________________________________ >We won't. Thanks for all your work in the past. Good luck to you. And to Red Hat I have one more thing to say: Buh bye! ### -- *Matt Phelps* *Information Technology Specialist, Systems Administrator* (Computation Facility, Smithsonian Astrophysical Observatory) Center for Astrophysics | Harvard & Smithsonian 60 Garden Street | MS 39 | Cambridge, MA 02138 email: mphelps at cfa.harvard.edu cfa.harvard.edu | Facebook <http://cfa.harvard.edu/facebook> | Twitter <http://cfa.harvard.edu/twitter> | YouTube <http://cfa.harvard.edu/youtube> | Newsletter <http://cfa.harvard.edu/newsletter>
Bernstein, Noam CIV USN NRL (6393) Washington DC (USA)
2020-Dec-16 02:13 UTC
[CentOS] CentOS 8 future
On Dec 15, 2020, at 7:41 PM, Johnny Hughes <johnny at centos.org<mailto:johnny at centos.org>> wrote: $250K is not even close. That is one employee, when you also take into account unemployment insurance, HR, medical insurance etc. now multiply that by 8. Now, outfit those 8 employees to work from home .. all over the world, different countries, different laws. Every package that ends up in a RHEL point release is in Stream at some point, right? While I can certainly believe that the cost for the entire CentOS effort is much more than $250K, dropping CentOS point releases just means not gathering the particular versions that ended up in the corresponding RHEL point release. Even for someone outside of CentOS, it sounds as simple as constantly downloading everything that's released in Stream (since apparently old rpm revisions won't stay in the CentOS repo), then looking at which versions made it into the RHEL point release, and copying just those to a repo for update. Am I missing some complex step?
On Tue, Dec 15, 2020 at 7:41 PM Johnny Hughes <johnny at centos.org> wrote:> $250K is not even close. That is one employee, when you also take into > account unemployment insurance, HR, medical insurance etc. now multiply > that by 8. Now, outfit those 8 employees to work from home .. all over > the world, different countries, different laws.I'm genuinely curious about something, and this is mostly academic since it's probably the subject of proprietary discussions within RedHat. Presumably, RedHat had a build pipeline for RHEL that worked well for them, by supplying alpha/beta releases of point releases to their customers and giving them time to "cook" before releasing those point releases into production. Why would RedHat invest millions more in buying the CentOS process just to have CentOS act as the beta?
Hi Johnny,> $250K is not even close. That is one employee, when you also take into > account unemployment insurance, HR, medical insurance etc. now multiply > that by 8. Now, outfit those 8 employees to work from home .. all over > the world, different countries, different laws. > > .. THEN buy 30 machines minimum (servers, not workstations) for > building and testing, buy a service contract for those 30 machines, host > the bandwidth required to sync out to 600 worldwide servers. > > We need all the CI machines .. that is a bunch of blade servers for > that. They need service contacts too.I don't doubt your numbers, they sound perfectly reasonable to me. On the other hand: How many of the employees will be laid off or reallocated now that CentOS point releases are no longer published? How many of the servers will be shut down, how many service contracts will be cancelled? What's your estimate of the reduction in bandwidth that will be saved by replacing point releases by a stream of releases with more frequent updates?> In any event it doesn't matter. The decision is made. If people don't > want to use CentOS Stream, then don't. The decision is not changing.Too bad. I've just completed a migration of about 30 servers from CentOS 6 to CentOS 8, expecting to get another 9 years of lifetime out of that (substantial) work. Now I have one year left of that, in which I need to plan what to do. One option is to go with the flow and switch to Stream, but I must admit that it's not my favourite one. Rocky, Lenix or maybe Springsdale would be the next best guesses. But given the fact that I migrated the whole setup process to Ansible it might be a good idea to jump off the cliff and switch to Debian or FreeBSD. As I said, I have one year left which I plan to use for evaluation of options. Two of my big customers will definitely not have that range of options. One of them is a RHEL shop with a tendency to try Debian, and last week they strongly thought about leaving the RHEL space entirely. The FOSS team there had made substantial effort over the last year to get CentOS on the list of company-approved operating systems (currently that's only RHEL and Debian), and now that work has gone down the drain completely. You can imagine how they feel now. The other one is stuck with RHEL-based distributions (Oracle, you know) - but they consider switching to OEL with support as well. At least they'll get rid of the hassle with the RHN that way, which can be a pain in the backside. I doubt those two are the only ones. My guess is this decision will backfire big time. I would love to stand corrected in one year's time, because I really like the RHEL way of doing things. Or rather, I liked it. Until last week. Still a great set of products, but the trustworthiness of Red Hat has taken a big hit for me, and for my customers as well. Anyway, thank you and the rest of the CentOS team for all the great work you've done and are doing. It is appreciated, and it will not be forgotten. Peter.