On 06/21/2014 03:31 PM, Roger Pena Escobio wrote:
Once again If the goal is to point to the fact centos 6.4 is different than rhel6.4 after 6.5 is out, then make something right when the differentiation start and not before You can release a new centos release package for 6.4 or you can do something on the repo, that make it clear to client, that something changed and updates do not work anymore unless the SA does some concious local change to make it not complain again, will that break some auto update on those clientes? Yes, but that would be desired to make it clear that something is requiered from the local SA in order to keep the box secure
But, I still see a big plus by having a release like mayor.minor.date, it is more flexible without breaking the old way , both side wins because a client could see/notice that their 7.0.x is too old when it was getting refreshed every now and then (asuming there will be respins more oftens that rh releases)
Ok, let me go further in a mental exercise, let say we make the change proposed and the .minor number is changed to .date of release. Lets imaging a year from that release there is serious vulnerability that put the world to check "am I vulnerable?". Manager or lazy SA check that we are running 7.20140701, they are not familiar with, they might even remember that was created when rhel7.0 was released but to be sure they go to a wiki and confirm that indeed that release of centos is the rebuild of rhel7.0 but they probably also notice that there are newer resping of centos based on newer releases of rhel, and probably they might notice that what they are running is no supported anymore raising more questions to them. What this might have different if we keep things as they are right now ? Probably because of what you are saying , that hypotetical person might hope/expect/ that centos is also following what redhat is doing and a fix will come in there way but at that point updates will not work unless conciously a manual change was performed and a the only update available might be a new centos release saying it is not supported
Make sense ? Again, having major.minor.date format looks like a compromise that have the best of both worlds
Thank roger
Sent from Yahoo! Mail on Android
*From: * Johnny Hughes johnny@centos.org; *To: * centos-devel@centos.org; *Subject: * Re: [CentOS-devel] CentOS 7 and release numbering *Sent: * Sat, Jun 21, 2014 10:42:26 AM
On 06/21/2014 05:00 AM, Ron Yorston wrote:
Johnny Hughes wrote:
What better way to communicate that they are not standalone but are all only part of the MAJOR release and a POINT IN TIME part of that major release than to name them "<MAJOR RELEASE>.<POINT IN TIME>" ?
The current scheme represents <POINT IN TIME> as an integer that starts from zero and increments with each minor release.
I remain unconvinced that a YYMM representation of <POINT IN TIME> is any better.
It is not really better at conveying time, no. It is the same at conveying the time.
Where it is better is in denoting that Red Hat is doing things inside the 6.4 tree (again, just following the above example) while CentOS does not do those things inside our 6.4 tree after we release 6.5. We can't do them, even if we want to as we don't have the sources.
That is my whole point .. we need a way to convey a similarity and one point, while not being similar always. Having the exact same name does not convey that.
How do you suggest we do that and not ignore that there are potential differences after we move to the next point release? Do we just ignore that part?
Everything on this list that is newer than 2013-11-20 is in the RHEL 6.4 tree ... we don't and can't release any of it for our 6.4:
https://rhn.redhat.com/errata/rhel-server-6.4.aus-errata.html
So our 6.4 tree is now significantly divergent from the Red Hat 6.4 tree, and our 6.4 tree is in the vault and not live anymore ... don't we have an obligation to our users to make sure they understand that there are differences?
UserA has some software that only works with 6.4 .. he sees CentOS-6.4 in the vault and grabs that to use with his software. He can't upgrade to 6.5 because it will break his software. Staying on our 6.4 tree will leave UserA vulnerable with security issues. If he is instead on the Red Hat 6.4 tree, he is still going to be able to get updates. Do we not have any obligation to change our numbering so that UserA can more easily tell this hugely major difference?
We don't really have the upstream point releases, we have different point releases. We release the main line CentOS-5, CentOS-6, and CentOS-7 ... we do point in time respins of ISOs and install trees, Red Hat does all this and a bunch more things also inside point releases. These two things are not EXACTLY the same ever, but they are very similar for one 6 to 8 month "period of time" (while they are OUR active release and Red Hat's active release) and they become increasing divergent after that point in time. That is what I am trying to convey here. Some people will argue that people have to pay for that other REd Hat 6.4 tree ... sure they do. They also have to pay the initial Red Hat 6.4 tree, they have to pay for everything there, thats how it works.
Everyone here thinks that we should just leave the point releases as is, knowing that now Red Hat is doing completely different things inside point releases and that we don't have an obligation to point out the differences?
I "yum update" moves you to newer version, then it is all the SAME as today. Updating WILL update that system beyond support.
If we need to reproduce point-in-time releases, then lets create base + updates repositories for every minor release (6.4 for example) and update it with only security fixes.
If we do not want to change the way CentOS is updated, then all we are saying to people is "We do not WANT to be same as RHEL, so if you want RHEL then buy it" which will brake a trust people have in CentOS binary compatibility with RHEL.