On 18 September 2013 11:45, "Jóhann B. Guðmundsson" <johannbg@gmail.com> wrote:
On 09/18/2013 05:37 PM, Kevin Fenzi wrote:
EPEL is a valid subproject/SIG of Fedora, and any changes we propose
need to take it into account, just like any other part of Fedora we are
currently supporting.

Well can we then make them clean up their spec file changes and keep them in separated branch?

Wat? They are in their own branches. Now if you are saying that maintainers should not have %{epel} in a fedora spec file.. well that is between you and FESCO or you and the maintainer. 


To get back to the actual subject of this thread, the current status of
running our own bugzilla is that we decided that we don't currently
have resources or desire to do so, and wanted to try and work with
existing bugzilla maintainers to try and address our concerns.

If there's things that change, we can change our plan.

So, constructive things to do moving forward:

* Clearly enumerate the issues with the current bugzilla and we can ask
   the bugzilla team to see if they can address them. If they do, then
   things will be better for us all. If they don't, we will know what
   items are causing problems and we need to specifically address in any
   solution we run ourselves. The wiki page is a good place to add/note

* Convince us that something has changed that would make running our
   own more attractive. For me at least, those would include: More
   people committed to helping, people with lots of bugzilla, perl or db
   knowledge committed to helping, some vastly better option than
   bugzilla appears, bugzilla itself becomes easier/better for our needs
   upstream, promise of more hardware to run our own on, serious
   issues unaddressed by current bugzilla admins, etc etc.

Just my 2 cents.

So the freedom for us to administrate and hack on our own instance is not good enough and you play the resource card?

1) Any bugzilla would require a lot of hardware/software. The current bugzilla runs with multiple front ends (2-4) and multiple back end database servers (somewhere between 7 and 10). We are one of the largest users of the Red Hat bugzilla so we would not be needing anything less because they aren't there for storage as much as scalability so that is a starting project price of $70->$120k not including power, cooling, storage, bandwidth and maintenance. (fast storage may make it much more). From talks with other large sites using Jira, Mantis, etc this will not change on which bug system we use because it is the nature of the number of bugs, lookups, updates, etc. If Fedora QA is interested in it, we can look at requesting from Red Hat that in the next fiscal year. 

2) The large bug bases require at least 2 full time people dealing with them. Most volunteers are part-time people who tend to start them up, burn out, get replaced with new volunteers who reimplement, etc. Volunteers are useful if a full time people are around.

3) We would need a complete bug day for any bug system we would use because existing bugs rely on lots of internal sql code which would be stuff Johann wants to remove for either slowness or not Fedora specific calls. Removing them might lower the number of scaling systems but most of the bug people have said you just replace them quickly with new items which remove any savings.

Final point,  EPEL is not just for RHEL. EPEL is what brings a lot of people into Fedora because they see a need for a package they want in RHEL and find out that they need to help it in Fedora before they can get it in EPEL. Also the number of systems using EPEL is 10x the number of Fedora users. So trying to get rid of EPEL is cutting off ones nose to spite ones face.  If you do not like Red Hat is the primary sponsor for Fedora, then I am sorry, but there isn't anything that I or anyone else here on this list can do.

Stephen J Smoogen.