============ Background ============
So my time at FOSDEM was pretty much the following. Walk into the hall where the Fedora/CentOS tables were. Get met by someone who worked on or in EPEL. Try to have a conversation and find that it was too packed and loud there. Walk over to the coffee area and buy the person a coffee and listen. Finish conversation and walk back over to the table.. [take a break every now and then to use the restroom from drinking a 16 oz tea every 30 minutes.] This was a lot of conversations and they did all blur together after a while. [Using a technique I learned from Centos' Karanbir Singh at FOSDEM, I will bring a small notebook to the next conference and write down a problem per page per person. Then I can go back the next year and see if the problems are still there.]
========== Promises ==========
While they all did blur, there were many similar questions from people that came up:
* What are the promises that EPEL makes to its users? * Do those promises make sense? [Do we need to keep a package the same for 5+ years?] * Do we actually keep those promises? * Who thinks we should be keeping these promises that various people didn't feel were needed? * What promises can we make that make sense and that after ~8 years we feel we can reasonably keep? * Who do we need to make these promises to?
As I said in the previous post, a lot of people felt that we had made promises of fixing and updating things in EPEL over the last 3 years but nothing really seemed to have changed. There are a lot of reasons for that (as in people say things and after a 20 hour flight have no idea what they said to looking at the promise and going "ok that is fine but who is going to do the work?" to "are we even allowed to change this?"
The last point is to me an undercurrent of a lot of conversations in and around EPEL in the last 3 years. There is the feeling that we were chartered nearly 10 years ago to do certain things, and we should keep doing them because people depend on that. And it is true that some people do depend on it, but it seems increasingly that they either don't show up to help or are a smaller set of people than we realize.
These questions and the problems above was a long discussion between Brian Stinson and me on Sunday evening. One of the circles that kept coming up was "Does EPEL need to be rechartered?" EPEL is a very old Fedora SIG, and it has gone through several incarnations in the last 8 years from when it had yearly elections for board members to when it ran without any leadership to its recent experimentation with a new steering committee. However during that time there have been many rules which it has tried to keep:
1. Do not do disruptive upgrades. Try to backport security fixes or do upgrades which do not change API/ABI issues. 2. Be prepared to support the same package for the life of an Enterprise Linux release. 3. Never replace or conflict with packages in the base Enterprise Linux. 4. They should never require manual updates or changes between updates. 5. Follow Fedora rules for bundling, licensing, and similar things.
There are some promises some people feel like we have made also:
1. Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ] 2. Packages will work from EL-X.0 to EL-X.N. [EPEL will rebuild packages regularly to deal with any ABI items] 3. Packages are built against all of an Enterprise Linux 'base' (EG whatever is in CentOS/Scientific Linux base).
In many cases those promises aren't kept now or when they are attempted to be kept are impossible to do.
* Most software which uses a database back end requires schema updates ranging from dump/reload to long running change processes. * A lot of software rebases itself internally to the point that trying to backport a security fix usually requires a lot of coding skill in the package. * The Enterprise Linux lifetime has increased from 5 years to 10 years since EPEL started. While people were interested in supporting a package for 3-4 years.. doing so for 10 years is too much for a volunteer position. * The Enterprise Linux is not the same as it was in EL-3/4/5 where software versions were rarely 'rebased' to a newer release. Instead software would be recoded to work in the older software as much as possible. Currently in EL-7, the desktop and other parts of the OS were greatly updated between 7.1 and 7.2 with the idea that this will be the norm for all releases in order to keep the product relevant to the majority of users.
From the above, 3 of the 5 promises are either too much to keep or haven't been kept in a long time.
============= What to Do? =============
Because we haven't been able to keep many promises, but feel like we should there seem to be a couple of options ahead: 1. Ignore and do nothing. I don't like the status-quo but I think it needs to be listed as something that can happen. 2. Try to keep the original charter and be a lot more picky about what is in EPEL so that we can meet those promises. 3. Figure out what we can do, and go to the appropriate Fedora body to recharter ourselves to meet those more reasonable goals.
Now number 3 may seem to be busy work, but I have found for the conservative minded enterprise maintainer, it is very hard to give up doing something even if you are failing until you are told you can stop trying. It also makes it clearer to future people working on EPEL in 2024 what they are trying to solve and how to change when the world of 2024 needs different solutions.
============================= What would be in my charter =============================
1. EPEL is run by the EPEL Steering Committee. They act as sub-equivalents to the Fedora Extras Steering Committee, Fedora Packaging Committee and Fedora Release Engineering. 2. Packages in EPEL will never replace or conflict with packages in the base Enterprise Linux. 3. Packages in EPEL will follow Fedora rules for bundling, licensing, and similar things. Exceptions to this will exist and will be documented in appropriate place. 4. EPEL release engineering can set up systems to rebase packages in regular intervals. [There are multiple ways of doing this that I will try to outline in other proposals this week.] 5. The only promise of a lifetime of a package is between dot releases of the Enterprise Linux. Maintainers should make an honest effort to support a package for the "life" of a sub-release (6.8->6.9) but no longer. 6. Packages in a dot release will not disappear during that time. There is no promise after that time. [If EPSco approves of a method that packages are regularly archived, then users can use that as a simpler method to get old packages.] 7. EPEL only supports the current dot release. If a package in EL-6.now works on 6.0.z that is great. If it doesn't, it is up to the user (and not the maintainer or EPEL) to deal with it. [Again if an archive method is setup, then those older versions would just be in something like /pub/archives/epel/epel-6.8/ or some similar setup]
Those are my starting points for a rechartered EPEL. Please join in the discussion on the EPEL-devel mailing list.
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
...snip...
However during that time there have been many rules which it has tried to keep:
- Do not do disruptive upgrades. Try to backport security fixes or do upgrades which do not change API/ABI issues.
- Be prepared to support the same package for the life of an Enterprise Linux release.
- Never replace or conflict with packages in the base Enterprise Linux.
- They should never require manual updates or changes between
updates. 5. Follow Fedora rules for bundling, licensing, and similar things.
There are some promises some people feel like we have made also:
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
- Packages will work from EL-X.0 to EL-X.N. [EPEL will rebuild packages regularly to deal with any ABI items]
Well, not regularly, but when there is a problem. With one of the recent 7.x releases there were a number of packages that needed rebuild. We didn't do a very nice job of it, but in the end we collected the bugs, rebuilt all those things and pushed the updates out. It would definitely be better to automate that and find these asap instead of after the fact with bugs.
- Packages are built against all of an Enterprise Linux 'base' (EG whatever is in CentOS/Scientific Linux base).
I thought we were always clear that we built against RHEL. But perhaps not.
In many cases those promises aren't kept now or when they are attempted to be kept are impossible to do.
- Most software which uses a database back end requires schema updates ranging from dump/reload to long running change processes.
- A lot of software rebases itself internally to the point that trying to backport a security fix usually requires a lot of coding skill in the package.
- The Enterprise Linux lifetime has increased from 5 years to 10 years since EPEL started. While people were interested in supporting a package for 3-4 years.. doing so for 10 years is too much for a volunteer position.
- The Enterprise Linux is not the same as it was in EL-3/4/5 where software versions were rarely 'rebased' to a newer release. Instead software would be recoded to work in the older software as much as possible. Currently in EL-7, the desktop and other parts of the OS were greatly updated between 7.1 and 7.2 with the idea that this will be the norm for all releases in order to keep the product relevant to the majority of users.
From the above, 3 of the 5 promises are either too much to keep or haven't been kept in a long time.
Yeah, those I all agree with.
============= What to Do? =============
Because we haven't been able to keep many promises, but feel like we should there seem to be a couple of options ahead:
- Ignore and do nothing. I don't like the status-quo but I think it needs to be listed as something that can happen.
- Try to keep the original charter and be a lot more picky about what is in EPEL so that we can meet those promises.
- Figure out what we can do, and go to the appropriate Fedora body to recharter ourselves to meet those more reasonable goals.
Now number 3 may seem to be busy work, but I have found for the conservative minded enterprise maintainer, it is very hard to give up doing something even if you are failing until you are told you can stop trying. It also makes it clearer to future people working on EPEL in 2024 what they are trying to solve and how to change when the world of 2024 needs different solutions.
I'm really not sure talking to another fedora body is needed. I really suspect if we came up with a new charter and all (all the people doing work on EPEL) agreed to it, we could just do it. FESCo or the council is likely to just say "sure, thats fine, do whatever". But we could do so if everyone wants...
============================= What would be in my charter =============================
- EPEL is run by the EPEL Steering Committee. They act as sub-equivalents to the Fedora Extras Steering Committee, Fedora Packaging Committee and Fedora Release Engineering.
(No extras anymore. ;)
- Packages in EPEL will never replace or conflict with packages in the base Enterprise Linux.
What does that mean? If we want it to be what we have now:
"Packages in EPEL will never replace or conflict with packages in the RHEL channels we build EPEL against. Adding or removing channels is done by the steering committee"
- Packages in EPEL will follow Fedora rules for bundling, licensing, and similar things. Exceptions to this will exist and will be documented in appropriate place.
- EPEL release engineering can set up systems to rebase packages in regular intervals. [There are multiple ways of doing this that I will try to outline in other proposals this week.]
Yeah, 'regular' is a bit too wishy washy.
- The only promise of a lifetime of a package is between dot releases of the Enterprise Linux. Maintainers should make an honest effort to support a package for the "life" of a sub-release (6.8->6.9) but no longer.
- Packages in a dot release will not disappear during that time. There is no promise after that time. [If EPSco approves of a method that packages are regularly archived, then users can use that as a simpler method to get old packages.]
I think these might be the most controversial bits.
- EPEL only supports the current dot release. If a package in EL-6.now works on 6.0.z that is great. If it doesn't, it is up to the user (and not the maintainer or EPEL) to deal with it. [Again if an archive method is setup, then those older versions would just be in something like /pub/archives/epel/epel-6.8/ or some similar setup]
Yeah, I think we have always said we only support the current release, but good to spell it out.
Those are my starting points for a rechartered EPEL. Please join in the discussion on the EPEL-devel mailing list.
I think we need to consider the window between rhel release and centos release. Do we wait? Do we not wait? do we test?
A good start anyhow, thanks smooge.
kevin
On February 18, 2016 11:56:54 AM PST, Kevin Fenzi kevin@scrye.com wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
...snip...
- Packages in EPEL will never replace or conflict with packages in the base Enterprise Linux.
What does that mean? If we want it to be what we have now:
"Packages in EPEL will never replace or conflict with packages in the RHEL channels we build EPEL against. Adding or removing channels is done by the steering committee"
Even that's not always feasible, like when RHEL suddenly started including the RHGS-only gluster client.
On Thu, 18 Feb 2016 12:11:28 -0800 Joe Julian joe@julianfamily.org wrote:
On February 18, 2016 11:56:54 AM PST, Kevin Fenzi kevin@scrye.com wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
...snip...
- Packages in EPEL will never replace or conflict with packages in the base Enterprise Linux.
What does that mean? If we want it to be what we have now:
"Packages in EPEL will never replace or conflict with packages in the RHEL channels we build EPEL against. Adding or removing channels is done by the steering committee"
Even that's not always feasible, like when RHEL suddenly started including the RHGS-only gluster client.
True. I recall someone getting on my case about the 'never' when there's often overlap because RHEL does something and we don't notice it right away. So, how about:
"EPEL strives to provide packages that do not conflict with or replace the packages in the RHEL channels that it builds against. From time to time such packages may be in the collection, but will be removed in a timely manner"
There's 0 chance we can always not be conflicting unless we somehow knew in advance any changes and had things ready for them when they changed.
kevin
Joe Julian wrote:
Even that's not always feasible, like when RHEL suddenly started including the RHGS-only gluster client.
Well, that's a "complex" argument.
Initially: The Storage Gluster client was in a RHEL _child_ channel at first, which was _not_ "recognized" as being "off-limits" by EPEL.
Then: The Storage Gluster client was moved into the RHEL base channel itself with a specific RHEL 5 and 6 Update, so suddenly "conflicting" overnight.
So it's more of the definition of "conflict" is a "complext" argument.
E.g., if a customer had Storage Gluster, it was _always_ conflicting, even when in the _child_ channel, even if _not_ "recognized" by EPEL as a "conflict."
Which goes to ...
Kevin Fenzi wrote:
True. I recall someone getting on my case about the 'never' when there's often overlap because RHEL does something and we don't notice it right away.
Back through 2011, every Red Hat major account "trusted" EPEL not to conflict.
But by 2013, virtually all Red Hat major accounts (at least all I was exposed to) _outlawed_ EPEL from being enabled at all, because of the endless conflicts with countless, various Red Hat products.
So ... today, in 2016, it doesn't really matter. Back in 2012, I was more worried about EPEL becoming something customers cannot enable. But that's now the reality.
I.e., Most Red Hat customers _never_ enable EPEL on systems, and they just "temporary enable" on very "select" systems for niche purposes.
E.g., if you are a Red Hat customer who has any Red Hat product that is not "protected" from conflicts with EPEL, you will have a mess if you enable it ... so none of them do.
Just how it is. No sense in reversing things now, unlike back in 2012. Red Hat customers, especially major accounts with at least a few add-ons, just cannot flirt with using EPEL.
It's become a dev and niche repo for them. Again, that's fine now, in 2016. If there was a way to avoid this, this was something that should have been done back in 2012. So I don't see any need to change course, today.
-- bjs
I was trying to reply to everything inline, but being the responsible posters ya'll are, you trimmed and everything leaving context harder to just jump in on :)
Way back in the day when I had a lot more time to spend on EPEL, I loved it. There were a few assumptions made then that I think don't hold up now.
1. 5 year lifecycle. 5 years is doable in a stretch by some maintainers. Now, it's 10 or 13 years depending on how you slice your RHEL support. That's too long for any volunteer. 2. Upstreams move a heck of a lot faster. As users/developers started trusting update mechansims more and speed/agility were seen as good rather than anti-stable, upstreams started moving. That means a LONG support release from some upstream players might be 18 months vs 4 years in the mid 2000s.
Those two assumptions not holding up are enough of a reason to recharter.
One thing I wished we had done early on was attract EPEL maintainers or separate from Fedora where desired. I found that lots of people only wanted to maintain packages in Fedora. They didn't care about EPEL at all. I found that people who really cared about the EL stack, didn't care about Fedora, but to help with EPEL, you had to be a Fedora person. I am not sure on the status of all that any more (cause I'm delinquent and way less active than I probably should be), but if that's still the case, could that be fixed in any way? Now that CentOS is under the RH umbrella it seems like there could be an enterprise community contributor worflow/org/group thing. (Apologies if there already is, again, I'm just popping my head back up after a while away).
<snip>
But by 2013, virtually all Red Hat major accounts (at least all I was exposed to) _outlawed_ EPEL from being enabled at all, because of the endless conflicts with countless, various Red Hat products.
So ... today, in 2016, it doesn't really matter. Back in 2012, I was more worried about EPEL becoming something customers cannot enable. But that's now the reality.
FWIW, I haven't seen this reaction, but a few anecdotes are also not
data. I do think as the OS has become more commoditized and the layered products have become the differentiator, this is more of a truism though, and should be planned for. Most people I see now mirror parts of EPEL and their own packages and other stuff and slam that all together into their package sets.
On February 18, 2016 7:08:34 PM PST, Michael Stahnke stahnma@puppetlabs.com wrote:
I was trying to reply to everything inline, but being the responsible posters ya'll are, you trimmed and everything leaving context harder to just jump in on :)
Way back in the day when I had a lot more time to spend on EPEL, I loved it. There were a few assumptions made then that I think don't hold up now.
- 5 year lifecycle. 5 years is doable in a stretch by some
maintainers. Now, it's 10 or 13 years depending on how you slice your RHEL support. That's too long for any volunteer. 2. Upstreams move a heck of a lot faster. As users/developers started trusting update mechansims more and speed/agility were seen as good rather than anti-stable, upstreams started moving. That means a LONG support release from some upstream players might be 18 months vs 4 years in the mid 2000s.
A large part of that, too, has been the widespread implementation of CI testing. There are fewer and fewer packages that consider any branch "unstable" anymore.
Those two assumptions not holding up are enough of a reason to recharter.
One thing I wished we had done early on was attract EPEL maintainers or separate from Fedora where desired. I found that lots of people only wanted to maintain packages in Fedora. They didn't care about EPEL at all. I found that people who really cared about the EL stack, didn't care about Fedora, but to help with EPEL, you had to be a Fedora person. I am not sure on the status of all that any more (cause I'm delinquent and way less active than I probably should be), but if that's still the case, could that be fixed in any way? Now that CentOS is under the RH umbrella it seems like there could be an enterprise community contributor worflow/org/group thing. (Apologies if there already is, again, I'm just popping my head back up after a while away).
<snip>
But by 2013, virtually all Red Hat major accounts (at least all I was exposed to) _outlawed_ EPEL from being enabled at all, because of the endless conflicts with countless, various Red Hat products.
So ... today, in 2016, it doesn't really matter. Back in 2012, I was more worried about EPEL becoming something customers cannot enable. But that's now the reality.
FWIW, I haven't seen this reaction, but a few anecdotes are also not
data. I do think as the OS has become more commoditized and the layered products have become the differentiator, this is more of a truism though, and should be planned for. Most people I see now mirror parts of EPEL and their own packages and other stuff and slam that all together into their package sets.
epel-devel mailing list epel-devel@lists.fedoraproject.org http://lists.fedoraproject.org/admin/lists/epel-devel@lists.fedoraproject.or...
On 18 February 2016 at 12:56, Kevin Fenzi kevin@scrye.com wrote:
- Packages are built against all of an Enterprise Linux 'base' (EG whatever is in CentOS/Scientific Linux base).
I thought we were always clear that we built against RHEL. But perhaps not.
Sorry what I meant that it is built against all of RHEL not just a couple of channels. Again this isn't a promise we ever made, but one people assume we have made (and get surprised when they find out differently).
============= What to Do? =============
Because we haven't been able to keep many promises, but feel like we should there seem to be a couple of options ahead:
- Ignore and do nothing. I don't like the status-quo but I think it needs to be listed as something that can happen.
- Try to keep the original charter and be a lot more picky about what is in EPEL so that we can meet those promises.
- Figure out what we can do, and go to the appropriate Fedora body to recharter ourselves to meet those more reasonable goals.
Now number 3 may seem to be busy work, but I have found for the conservative minded enterprise maintainer, it is very hard to give up doing something even if you are failing until you are told you can stop trying. It also makes it clearer to future people working on EPEL in 2024 what they are trying to solve and how to change when the world of 2024 needs different solutions.
I'm really not sure talking to another fedora body is needed. I really suspect if we came up with a new charter and all (all the people doing work on EPEL) agreed to it, we could just do it. FESCo or the council is likely to just say "sure, thats fine, do whatever". But we could do so if everyone wants...
============================= What would be in my charter =============================
- EPEL is run by the EPEL Steering Committee. They act as sub-equivalents to the Fedora Extras Steering Committee, Fedora Packaging Committee and Fedora Release Engineering.
(No extras anymore. ;)
- Packages in EPEL will never replace or conflict with packages in the base Enterprise Linux.
What does that mean? If we want it to be what we have now:
"Packages in EPEL will never replace or conflict with packages in the RHEL channels we build EPEL against. Adding or removing channels is done by the steering committee"
And this is where I made things worse by conflicting with channels versus "entire OS"
- Packages in EPEL will follow Fedora rules for bundling, licensing, and similar things. Exceptions to this will exist and will be documented in appropriate place.
- EPEL release engineering can set up systems to rebase packages in regular intervals. [There are multiple ways of doing this that I will try to outline in other proposals this week.]
Yeah, 'regular' is a bit too wishy washy.
Well this was more of a rule. The exact procedures of how this is done would be in a separate document so that it can be changed easily.
- The only promise of a lifetime of a package is between dot releases of the Enterprise Linux. Maintainers should make an honest effort to support a package for the "life" of a sub-release (6.8->6.9) but no longer.
- Packages in a dot release will not disappear during that time. There is no promise after that time. [If EPSco approves of a method that packages are regularly archived, then users can use that as a simpler method to get old packages.]
I think these might be the most controversial bits.
Probably but I would prefer to either have them one way or another :). I would prefer not to have the "where is my N?" during the decay days of EPEL-6 that we are seeing in EL-5 as packages get EOL'd that people expected to be there.
- EPEL only supports the current dot release. If a package in EL-6.now works on 6.0.z that is great. If it doesn't, it is up to the user (and not the maintainer or EPEL) to deal with it. [Again if an archive method is setup, then those older versions would just be in something like /pub/archives/epel/epel-6.8/ or some similar setup]
Yeah, I think we have always said we only support the current release, but good to spell it out.
Those are my starting points for a rechartered EPEL. Please join in the discussion on the EPEL-devel mailing list.
I think we need to consider the window between rhel release and centos release. Do we wait? Do we not wait? do we test?
Well that is another promise some people think we make. They thought we waited until CentOS came out before the buildroots were changed but that isn't something I think we have ever done. I forgot to put that in. Part of this comes in with the procedures to be put in place. If it is too hard to deal with alt-arch it would be clearer that we said "We only support and build packages against the base RHEL release." Then we can rule out the window between rhel and centos.. and the altarch can come up with their own solution.
If we do want to have the alt-arch it comes into a bigger issue depending on how the alt-arch is implemented in EPEL. There seem to be multiple ways to do this that need information from Release Engineering to solve.
A good start anyhow, thanks smooge.
kevin
epel-devel mailing list epel-devel@lists.fedoraproject.org http://lists.fedoraproject.org/admin/lists/epel-devel@lists.fedoraproject.or...
Stephen John Smoogen wrote:
Sorry what I meant that it is built against all of RHEL not just a couple of channels. Again this isn't a promise we ever made, but one people assume we have made (and get surprised when they find out differently).
I can only imagine how difficult this gets for EPEL maintainers, even the ones with the RHEL Developer Suite subscription. There are just a lot of add-ons and channels to track, even ones not in the RHEL Dev.
As I've mentioned in years prior, the only way to prevent this is for the CentOS team to build everything, and that's a tall order, and only getting to be a bigger and bigger task by the day. So EPEL is really one of those things that Red Hat customers pull down, but don't enable ... until something specifically is needed, and then it's checked for conflicts, and only on a small subset of systems.
Non-Red Hat customers running CentOS usually have a much easier task of enabling EPEL, although there can still be some issues.
And this is where I made things worse by conflicting with channels versus "entire OS"
I don't think there is a perfect answer that satisfies all, and never will be. And definitely not 4-5 years into where EPEL is today, from where the Red Hat product line went more and more vertical than it ever was before.
It's one of those things that people have just accepted. Although it is good for maintainers to recognize that the EPEL build system is going to -- gasp -- build against Red Hat packages, and not CentOS. ;)
Probably but I would prefer to either have them one way or another :). I would prefer not to have the "where is my N?" during the decay days of EPEL-6 that we are seeing in EL-5 as packages get EOL'd that people expected to be there.
Maybe I'm throwing this out in ignorance ...
But maybe [Red Hat] Software Collections ([RH]SCL) offers a baseline for a "reasonable lifecycle" that should be a target? I.e., 3 years.
E.g., any software that makes it into EPEL will have a "soft target" of at least 3 years support?
Just a suggestion ... one that RHSCL users are familiar with. I think it's unreasonable to expect updates beyond that, or at least not without a rebase after 3 years.
As always, just my $0.02 ... from a Red Hat customer standpoint, not a community one.
On Thu, Feb 18, 2016 at 2:39 PM, Bryan J Smith b.j.smith@ieee.org wrote:
Stephen John Smoogen wrote:
Sorry what I meant that it is built against all of RHEL not just a couple of channels. Again this isn't a promise we ever made, but one people assume we have made (and get surprised when they find out differently).
I can only imagine how difficult this gets for EPEL maintainers, even the ones with the RHEL Developer Suite subscription. There are just a lot of add-ons and channels to track, even ones not in the RHEL Dev.
As I've mentioned in years prior, the only way to prevent this is for the CentOS team to build everything, and that's a tall order, and only getting to be a bigger and bigger task by the day. So EPEL is really one of those things that Red Hat customers pull down, but don't enable ... until something specifically is needed, and then it's checked for conflicts, and only on a small subset of systems.
Non-Red Hat customers running CentOS usually have a much easier task of enabling EPEL, although there can still be some issues.
And this is where I made things worse by conflicting with channels versus "entire OS"
I don't think there is a perfect answer that satisfies all, and never will be. And definitely not 4-5 years into where EPEL is today, from where the Red Hat product line went more and more vertical than it ever was before.
It's one of those things that people have just accepted. Although it is good for maintainers to recognize that the EPEL build system is going to -- gasp -- build against Red Hat packages, and not CentOS. ;)
Probably but I would prefer to either have them one way or another :). I would prefer not to have the "where is my N?" during the decay days of EPEL-6 that we are seeing in EL-5 as packages get EOL'd that people expected to be there.
Maybe I'm throwing this out in ignorance ...
But maybe [Red Hat] Software Collections ([RH]SCL) offers a baseline for a "reasonable lifecycle" that should be a target? I.e., 3 years.
E.g., any software that makes it into EPEL will have a "soft target" of at least 3 years support?
RHSCL is a non-starter where I work (and I imagine at other locations). 2-3 years of support just isn't enough to make it a worthwhile investment. For example, we just barely started upgrading to EL 7 and it will be years until a sizable portion of our machines have moved to EL 7 (we still have quite a few EL 5 boxes).
For most packages, leaving the version that's there alone is probably a decent solution, but for packages where security fixes continue to happen and are important, then an "official" upgrade path would be good. Maybe something like: 1) Current maintainer identifies upgraded version and reason for update (hopefully limited to security fixes or something along those lines and not just "I want a newer version"), 2) Notifies intention on mailing list, 3) Feedback from community, and 4) Perform upgrade
Something like this could potentially get maintainers to step up that are willing to continue maintaining the current version. But in most cases, I'm guessing this wouldn't happen but would give an official process and expected timeline for user users.
Dave Johansen wrote:
RHSCL is a non-starter where I work (and I imagine at other locations). 2-3 years of support just isn't enough to make it a worthwhile investment.
Well, there usually _is_ more than one (1) [RH]SCL per RHEL release. So it's more like 2-3 releases that "rebase" every 2+ years, for 6-7 years total of the RHEL release.
The idea here is that you have up to three (3) years to "rebase." ;)
Not that [RH]SCL is 3 years and that's it. I mean ... understand my intent here. Sustaining engineering is _not_ free, it's _not_ upstream, and people should be expected to rebase every 2-3 years, when it comes closer to Upstream.
Again, understand the "bigger picture" of my "suggestion."
For example, we just barely started upgrading to EL 7 ... (cut)
_So_ you will likely be on the _second_ [RH]SCL release for RHEL 7, instead of the _first_. So ... what's the problem?
So ... I have to now ask ... have you used [RH]SCL? ;)
I.e., Again, it's _not_ only 3 years of RHEL, but 3 years per release, with each RHEL releases getting 2-3 releases over 6-7 years.
Ergo ...
For most packages, leaving the version that's there alone is probably a decent solution, but for packages where security fixes continue to happen and are important, then an "official" upgrade path would be good. Maybe something like:
- Current maintainer identifies upgraded version and reason for update
(hopefully limited to security fixes or something along those lines and not just "I want a newer version"), 2) Notifies intention on mailing list, 3) Feedback from community, and 4) Perform upgrade
Which works out very well for something that rebases every 2-3 years ... like [RH]SCL.
Again, I wasn't trying to say "be like [RH]SCL," but more like, "Here's a Red Hat add-on that kinda already has this 'lifecycle' that Red Hat customers would understand."
-- bjs
On Thu, Feb 18, 2016 at 3:54 PM, Bryan J Smith b.j.smith@ieee.org wrote:
Dave Johansen wrote:
RHSCL is a non-starter where I work (and I imagine at other locations). 2-3 years of support just isn't enough to make it a worthwhile investment.
Well, there usually _is_ more than one (1) [RH]SCL per RHEL release. So it's more like 2-3 releases that "rebase" every 2+ years, for 6-7 years total of the RHEL release.
The idea here is that you have up to three (3) years to "rebase." ;)
Not that [RH]SCL is 3 years and that's it. I mean ... understand my intent here. Sustaining engineering is _not_ free, it's _not_ upstream, and people should be expected to rebase every 2-3 years, when it comes closer to Upstream.
Again, understand the "bigger picture" of my "suggestion."
For example, we just barely started upgrading to EL 7 ... (cut)
_So_ you will likely be on the _second_ [RH]SCL release for RHEL 7, instead of the _first_. So ... what's the problem?
So ... I have to now ask ... have you used [RH]SCL? ;)
I.e., Again, it's _not_ only 3 years of RHEL, but 3 years per release, with each RHEL releases getting 2-3 releases over 6-7 years.
Ergo ...
SCL is an AWESOME idea, but for our organization, having to jump to the next version of an SCL every 2-3 years just isn't possible. We use RHEL because of its long support life cycle and starting to use the SCL breaks that idea. If SCL had a lifecycle more akin to RHEL (i.e. something like release half as often and support for twice as long), then it would be a different story.
For most packages, leaving the version that's there alone is probably a decent solution, but for packages where security fixes continue to happen and are important, then an "official" upgrade path would be good. Maybe something like:
- Current maintainer identifies upgraded version and reason for update
(hopefully limited to security fixes or something along those lines and
not
just "I want a newer version"), 2) Notifies intention on mailing list, 3) Feedback from community, and 4) Perform upgrade
Which works out very well for something that rebases every 2-3 years ... like [RH]SCL.
Again, I wasn't trying to say "be like [RH]SCL," but more like, "Here's a Red Hat add-on that kinda already has this 'lifecycle' that Red Hat customers would understand."
Sorry, what I was trying to get across was the idea of having a process that discouraged but allowed for upgrades in a controlled manner while giving users plenty of time to prepare for it.
On 18 February 2016 at 16:37, Dave Johansen davejohansen@gmail.com wrote:
On Thu, Feb 18, 2016 at 3:54 PM, Bryan J Smith b.j.smith@ieee.org wrote:
Dave Johansen wrote:
RHSCL is a non-starter where I work (and I imagine at other locations). 2-3 years of support just isn't enough to make it a worthwhile investment.
Well, there usually _is_ more than one (1) [RH]SCL per RHEL release. So it's more like 2-3 releases that "rebase" every 2+ years, for 6-7 years total of the RHEL release.
The idea here is that you have up to three (3) years to "rebase." ;)
Not that [RH]SCL is 3 years and that's it. I mean ... understand my intent here. Sustaining engineering is _not_ free, it's _not_ upstream, and people should be expected to rebase every 2-3 years, when it comes closer to Upstream.
Again, understand the "bigger picture" of my "suggestion."
For example, we just barely started upgrading to EL 7 ... (cut)
_So_ you will likely be on the _second_ [RH]SCL release for RHEL 7, instead of the _first_. So ... what's the problem?
So ... I have to now ask ... have you used [RH]SCL? ;)
I.e., Again, it's _not_ only 3 years of RHEL, but 3 years per release, with each RHEL releases getting 2-3 releases over 6-7 years.
Ergo ...
SCL is an AWESOME idea, but for our organization, having to jump to the next version of an SCL every 2-3 years just isn't possible. We use RHEL because of its long support life cycle and starting to use the SCL breaks that idea. If SCL had a lifecycle more akin to RHEL (i.e. something like release half as often and support for twice as long), then it would be a different story.
One of the issues that I am finding is that most software these days has only a 3 year lifetime at best.. if you want it to be longer you are going to pay for it either by maintaining it yourself or paying someone else. The work to try and keep software 'stable' over a 6 month lifetime seems to grow exponentially so that by the end of the 3 years it is 1000 times harder than it was in the beginning. Now the number of people who are wanting to work on this older software for free (or if they are doing it for themselves to share it for "free") is incredibly small. I expect that we are looking at 10 packages in EPEL maybe? So we are going to be looking at rebasing and updates every 2-3 years no matter what for the majority of software. Expecting it not to be like that is Cnut's tidal problem.
https://en.wikipedia.org/wiki/King_Canute_and_the_waves
On Thu, 18 Feb 2016 17:22:07 -0700 Stephen John Smoogen smooge@gmail.com wrote:
One of the issues that I am finding is that most software these days has only a 3 year lifetime at best.. if you want it to be longer you are going to pay for it either by maintaining it yourself or paying someone else. The work to try and keep software 'stable' over a 6 month lifetime seems to grow exponentially so that by the end of the 3 years it is 1000 times harder than it was in the beginning. Now the number of people who are wanting to work on this older software for free (or if they are doing it for themselves to share it for "free") is incredibly small. I expect that we are looking at 10 packages in EPEL maybe? So we are going to be looking at rebasing and updates every 2-3 years no matter what for the majority of software. Expecting it not to be like that is Cnut's tidal problem.
An additional issue with SCL's is that we have no guidelines on them.
While I suppose EPEL could add them with it's own guidelines, I would be against that given that Fedora couldn't finalize them.
kevin
On Thu, Feb 18, 2016 at 5:22 PM, Stephen John Smoogen smooge@gmail.com wrote:
On 18 February 2016 at 16:37, Dave Johansen davejohansen@gmail.com wrote:
On Thu, Feb 18, 2016 at 3:54 PM, Bryan J Smith b.j.smith@ieee.org
wrote:
Dave Johansen wrote:
RHSCL is a non-starter where I work (and I imagine at other locations). 2-3 years of support just isn't enough to make it a worthwhile investment.
Well, there usually _is_ more than one (1) [RH]SCL per RHEL release. So it's more like 2-3 releases that "rebase" every 2+ years, for 6-7 years total of the RHEL release.
The idea here is that you have up to three (3) years to "rebase." ;)
Not that [RH]SCL is 3 years and that's it. I mean ... understand my intent here. Sustaining engineering is _not_ free, it's _not_ upstream, and people should be expected to rebase every 2-3 years, when it comes closer to Upstream.
Again, understand the "bigger picture" of my "suggestion."
For example, we just barely started upgrading to EL 7 ... (cut)
_So_ you will likely be on the _second_ [RH]SCL release for RHEL 7, instead of the _first_. So ... what's the problem?
So ... I have to now ask ... have you used [RH]SCL? ;)
I.e., Again, it's _not_ only 3 years of RHEL, but 3 years per release, with each RHEL releases getting 2-3 releases over 6-7 years.
Ergo ...
SCL is an AWESOME idea, but for our organization, having to jump to the
next
version of an SCL every 2-3 years just isn't possible. We use RHEL
because
of its long support life cycle and starting to use the SCL breaks that
idea.
If SCL had a lifecycle more akin to RHEL (i.e. something like release
half
as often and support for twice as long), then it would be a different
story.
One of the issues that I am finding is that most software these days has only a 3 year lifetime at best.. if you want it to be longer you are going to pay for it either by maintaining it yourself or paying someone else. The work to try and keep software 'stable' over a 6 month lifetime seems to grow exponentially so that by the end of the 3 years it is 1000 times harder than it was in the beginning. Now the number of people who are wanting to work on this older software for free (or if they are doing it for themselves to share it for "free") is incredibly small. I expect that we are looking at 10 packages in EPEL maybe? So we are going to be looking at rebasing and updates every 2-3 years no matter what for the majority of software. Expecting it not to be like that is Cnut's tidal problem.
As mentioned several times already, EPEL means different things to different people. To me stability is significantly more important than having the latest and greatest, so for a fair number of packages that means "just leave things alone" and the "waves" are self inflicted.
But as was also mentioned, a lot of packages have jumped on the Chrome bandwagon with crazy release cycles and maintaining security fixes in those cases is basically impossible for a volunteer effort. From my perspective, having a policy that "discourages but allows updates" with a very deliberate and controlled process is the right model for EPEL.
Am 19.02.2016 um 16:53 schrieb Dave Johansen:
But as was also mentioned, a lot of packages have jumped on the Chrome bandwagon with crazy release cycles and maintaining security fixes in those cases is basically impossible for a volunteer effort. From my perspective, having a policy that "discourages but allows updates" with a very deliberate and controlled process is the right model for EPEL.
Sounds good to me.
Also as a sysadmin I dislike that stuff in EPEL can change at any time (whenever the maintainer can not support the old version anymore). If EPEL had some kind of "release" (e.g. tied to RHEL point releases) I'd like to see "most" incompatible upgrades happen at that time - with some kind of release notes where I can read about the changes before.
Ideally I have some time (e.g. a month or so) to schedule the upgrade and deal with any fall-out.
If that's possible with Stephen's rechartering I'm more than happy.
Felix
On Fri, Feb 26, 2016 at 2:38 PM, Felix Schwarz fschwarz@fedoraproject.org wrote:
Also as a sysadmin I dislike that stuff in EPEL can change at any time (whenever the maintainer can not support the old version anymore). If EPEL had some kind of "release" (e.g. tied to RHEL point releases) I'd like to see "most" incompatible upgrades happen at that time - with some kind of release notes where I can read about the changes before.
Ideally I have some time (e.g. a month or so) to schedule the upgrade and deal with any fall-out.
I'm curious, do you test updates that are going through epel-testing?
- Ken
Q: Do you understand the purpose of RHSCL PR EPEL? ;)
You want RHEL, but neither RHSCL nor EPEL is it, nor designed to be it.
That's why I said ...
EPEL's ideal aimpoint, when feasible, should be like RHSCL. ;)
-- bjs
DISCLAIMER: Sent from phone, please excuse any typos
On Sat, Feb 27, 2016 at 12:14 PM, Bryan J Smith b.j.smith@ieee.org wrote:
Q: Do you understand the purpose of RHSCL PR EPEL? ;)
You want RHEL, but neither RHSCL nor EPEL is it, nor designed to be it.
That's why I said ...
EPEL's ideal aimpoint, when feasible, should be like RHSCL. ;)
SCL is for parallel installable packages that already exist in the system (i.e. newer versions of gcc/Postgres/Python/etc). Some EPEL packages also provide that sort of thing (like the old cmake28 packages), but I don't believe that EPEL as a whole is aiming to do that more than just provide packages that aren't available in EL.
On 02/18/2016 01:39 PM, Bryan J Smith wrote:
Non-Red Hat customers running CentOS usually have a much easier task of enabling EPEL, although there can still be some issues.
Yes it is a bit odd to me that the free CentOS is in many respects easier to use than the paid RHEL.
Would be nice where support is needed to be able to install CentOS and then buy an RHEL support license but keep using CentOS.
On 02/18/2016 01:56 PM, Kevin Fenzi wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
One of the reasons why I run my own mirror for my server farm is so that I can rsync mirror EPEL _without_ doing a delete. Why? Because this happens to me _all_ the time:
Day 1: Install package from EPEL. Day 2: What do you mean the package isn't found? Look at main repo. Shit. Package was there yesterday! Dig through mailing list for an hour until I find a reason (some times I don't find a reason at all and have to dig through Fedora's Koji for one).
I get that there are a dozen valid reasons for a package to no longer be updated, but as long as it keeps working I don't understand why it has to be deleted. I have a couple dozen packages that I still install for my user base that haven't been in the main EPEL repo in a year.
My mirror for just 6 & 7 is ~100GB now. Small price for me since it doesn't delete which has saved my butt numerous times when a package suddenly disappears.
Thanks! ~Stack~
On 18 February 2016 at 14:13, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 01:56 PM, Kevin Fenzi wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
Software in EPEL follows many of the same rules as software in Fedora. If there isn't an active maintainer the software is removed. This is because a lot of people expect that the software is going to be getting updates and bug fixes when it is there. If it isn't there it is clear that it isn't getting any bug fixes.
While as a user this is a major pain in the butt, on the other side (maintainers and developers of the software) it is a major pain when the opposite occurs. Developers get complaints about software they no longer have any interest in and try to find someone to get rid of the old software. People who are maintainers of other packages get long hate emails about why is this software still in XYZ repository if no one is going to care about it. It burnt out a lot of the early repository people because they had made a package for someone at some point but really didn't have any care for it to be there any longer but all they were getting was crap for it being there.
So in Fedora and EPEL it is simple, you want a package that no one else wants.. you maintain the package that no one else wants.
One of the reasons why I run my own mirror for my server farm is so that I can rsync mirror EPEL _without_ doing a delete. Why? Because this happens to me _all_ the time:
Day 1: Install package from EPEL. Day 2: What do you mean the package isn't found? Look at main repo. Shit. Package was there yesterday! Dig through mailing list for an hour until I find a reason (some times I don't find a reason at all and have to dig through Fedora's Koji for one).
I get that there are a dozen valid reasons for a package to no longer be updated, but as long as it keeps working I don't understand why it has to be deleted. I have a couple dozen packages that I still install for my user base that haven't been in the main EPEL repo in a year.
My mirror for just 6 & 7 is ~100GB now. Small price for me since it doesn't delete which has saved my butt numerous times when a package suddenly disappears.
Thanks! ~Stack~
epel-devel mailing list epel-devel@lists.fedoraproject.org http://lists.fedoraproject.org/admin/lists/epel-devel@lists.fedoraproject.or...
On 02/18/2016 06:29 PM, Stephen John Smoogen wrote:
On 18 February 2016 at 14:13, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 01:56 PM, Kevin Fenzi wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
Software in EPEL follows many of the same rules as software in Fedora. If there isn't an active maintainer the software is removed. This is because a lot of people expect that the software is going to be getting updates and bug fixes when it is there. If it isn't there it is clear that it isn't getting any bug fixes.
While as a user this is a major pain in the butt, on the other side (maintainers and developers of the software) it is a major pain when the opposite occurs. Developers get complaints about software they no longer have any interest in and try to find someone to get rid of the old software. People who are maintainers of other packages get long hate emails about why is this software still in XYZ repository if no one is going to care about it. It burnt out a lot of the early repository people because they had made a package for someone at some point but really didn't have any care for it to be there any longer but all they were getting was crap for it being there.
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo? Community would know that it isn't maintained and yet the package wouldn't just disappear completely. I guess the difficult question would then be, how long is the package kept till it needs to be pruned? 1yr? 6mo? Still, it would be nice to give the user base the option to pull the packages they need out on a long enough scale that they have time to discover it with new builds.
I also wonder how many people have been bit by this. I know I have supplied packages out of my repos for others before, but it has only been a handful. It seems to have bitten me several times in the last year and all were for EL6, but (fingers crossed) I am done with the EL6 boxes in <2yrs with a full migration to either EL7 or Ubuntu LTS and won't need the EL6 repo any more. :-)
Thanks again! ~Stack~
On 02/19/2016 04:16 AM, ~Stack~ wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
Software in EPEL follows many of the same rules as software in Fedora. If there isn't an active maintainer the software is removed. This is because a lot of people expect that the software is going to be getting updates and bug fixes when it is there. If it isn't there it is clear that it isn't getting any bug fixes.
While as a user this is a major pain in the butt, on the other side (maintainers and developers of the software) it is a major pain when the opposite occurs. Developers get complaints about software they no longer have any interest in and try to find someone to get rid of the old software. People who are maintainers of other packages get long hate emails about why is this software still in XYZ repository if no one is going to care about it. It burnt out a lot of the early repository people because they had made a package for someone at some point but really didn't have any care for it to be there any longer but all they were getting was crap for it being there.
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo?
We could expand https://dl.fedoraproject.org/pub/archive/ for this purpose, after all
On 18 February 2016 at 19:16, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 06:29 PM, Stephen John Smoogen wrote:
On 18 February 2016 at 14:13, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 01:56 PM, Kevin Fenzi wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
Software in EPEL follows many of the same rules as software in Fedora. If there isn't an active maintainer the software is removed. This is because a lot of people expect that the software is going to be getting updates and bug fixes when it is there. If it isn't there it is clear that it isn't getting any bug fixes.
While as a user this is a major pain in the butt, on the other side (maintainers and developers of the software) it is a major pain when the opposite occurs. Developers get complaints about software they no longer have any interest in and try to find someone to get rid of the old software. People who are maintainers of other packages get long hate emails about why is this software still in XYZ repository if no one is going to care about it. It burnt out a lot of the early repository people because they had made a package for someone at some point but really didn't have any care for it to be there any longer but all they were getting was crap for it being there.
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo? Community would know that it isn't maintained and yet the package wouldn't just disappear completely. I guess the difficult question would then be, how long is the package kept till it needs to be pruned? 1yr? 6mo? Still, it would be nice to give the user base the option to pull the packages they need out on a long enough scale that they have time to discover it with new builds.
So every package ever signed is still in the koji build server if you know where to look for it. However I am hoping next week to write up some proposals that could accomplish this.
I also wonder how many people have been bit by this. I know I have supplied packages out of my repos for others before, but it has only been a handful. It seems to have bitten me several times in the last year and all were for EL6, but (fingers crossed) I am done with the EL6 boxes in <2yrs with a full migration to either EL7 or Ubuntu LTS and won't need the EL6 repo any more. :-)
A lot of people have been bit as it is nearly a daily occurrence in #epel on the irc server.
Thanks again! ~Stack~
epel-devel mailing list epel-devel@lists.fedoraproject.org http://lists.fedoraproject.org/admin/lists/epel-devel@lists.fedoraproject.or...
On 02/19/2016 08:13 AM, Stephen John Smoogen wrote:
On 18 February 2016 at 19:16, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 06:29 PM, Stephen John Smoogen wrote:
On 18 February 2016 at 14:13, ~Stack~ i.am.stack@gmail.com wrote:
On 02/18/2016 01:56 PM, Kevin Fenzi wrote:
On Tue, 16 Feb 2016 23:24:58 -0700 Stephen John Smoogen smooge@gmail.com wrote:
[snip]
- Packages will never disappear. [They don't disappear from Fedora 12 even if it is archived.. ]
To my understanding we never made this promise. We should try and communicate why it's NOT something we promise.
Could you elaborate on this please? I have asked before, but didn't really get an answer. I don't understand this.
Software in EPEL follows many of the same rules as software in Fedora. If there isn't an active maintainer the software is removed. This is because a lot of people expect that the software is going to be getting updates and bug fixes when it is there. If it isn't there it is clear that it isn't getting any bug fixes.
While as a user this is a major pain in the butt, on the other side (maintainers and developers of the software) it is a major pain when the opposite occurs. Developers get complaints about software they no longer have any interest in and try to find someone to get rid of the old software. People who are maintainers of other packages get long hate emails about why is this software still in XYZ repository if no one is going to care about it. It burnt out a lot of the early repository people because they had made a package for someone at some point but really didn't have any care for it to be there any longer but all they were getting was crap for it being there.
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo? Community would know that it isn't maintained and yet the package wouldn't just disappear completely. I guess the difficult question would then be, how long is the package kept till it needs to be pruned? 1yr? 6mo? Still, it would be nice to give the user base the option to pull the packages they need out on a long enough scale that they have time to discover it with new builds.
So every package ever signed is still in the koji build server if you know where to look for it. However I am hoping next week to write up some proposals that could accomplish this.
I also wonder how many people have been bit by this. I know I have supplied packages out of my repos for others before, but it has only been a handful. It seems to have bitten me several times in the last year and all were for EL6, but (fingers crossed) I am done with the EL6 boxes in <2yrs with a full migration to either EL7 or Ubuntu LTS and won't need the EL6 repo any more. :-)
A lot of people have been bit as it is nearly a daily occurrence in #epel on the irc server.
Thanks again! ~Stack~
My first comment is that I think we need to do better communication, as I don't think an announcement goes out to epel-users when a package is going to be retired.
My proposal would be to:
- When building against RHEL X.Y, the epel packages end up in epel/X.Y with a symbolic link of epel/X -> epel/X.Y. - The last two versions are in http://dl.fedoraproject.org/pub/epel, with older ones retired to http://archives.fedoraproject.org/pub/archive/epel
On 19/02/16 15:16, ~Stack~ wrote:
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo? Community would know that it isn't maintained and yet the package wouldn't just disappear completely. I guess the difficult question would then be, how long is the package kept till it needs to be pruned? 1yr? 6mo? Still, it would be nice to give the user base the option to pull the packages they need out on a long enough scale that they have time to discover it with new builds.
My suggestion would be for the life of the point release of the repos that's built against. Since the package is not going to be built against newer point releases of RHEL it is less likely to continue to work after RHEL moves to a new point release (say from 7.2 to 7.3). We could have an individual "retired" repo for each point release that would see the packages built against that release moved there. We would not necessarily need get rid of older retired repos, but just maintain a symbolic link to the latest one.
So for example if package foo was last built against RHEL 7.2 before it was retired, we could move it to the repo "epel-retired-7.2" and there would be a symbolic link for "epel-retired" that points to epel-retired-7.2 until RHEL 7.3 is released.
Peter
On 02/22/2016 09:06 PM, Peter wrote:
On 19/02/16 15:16, ~Stack~ wrote:
Thanks for replying. Makes sense. But what would the harm be to move a package into a separate "retired" repo? Community would know that it isn't maintained and yet the package wouldn't just disappear completely. I guess the difficult question would then be, how long is the package kept till it needs to be pruned? 1yr? 6mo? Still, it would be nice to give the user base the option to pull the packages they need out on a long enough scale that they have time to discover it with new builds.
My suggestion would be for the life of the point release of the repos that's built against. Since the package is not going to be built against newer point releases of RHEL it is less likely to continue to work after RHEL moves to a new point release (say from 7.2 to 7.3). We could have an individual "retired" repo for each point release that would see the packages built against that release moved there. We would not necessarily need get rid of older retired repos, but just maintain a symbolic link to the latest one.
So for example if package foo was last built against RHEL 7.2 before it was retired, we could move it to the repo "epel-retired-7.2" and there would be a symbolic link for "epel-retired" that points to epel-retired-7.2 until RHEL 7.3 is released.
In my opinion, that is a brilliant idea.
I like it! :-)
~Stack~
epel-devel@lists.fedoraproject.org