Response to "Getting Fedora Out of the If-Then Loop"

Josh Boyer jwboyer at
Sat Feb 20 01:10:56 UTC 2010

On Fri, Feb 19, 2010 at 11:18:42PM +0100, Christoph Wickert wrote:
>Am Freitag, den 19.02.2010, 16:40 -0500 schrieb Josh Boyer:
>> On Fri, Feb 19, 2010 at 10:12:12PM +0100, Christoph Wickert wrote:
>> >Am Freitag, den 19.02.2010, 10:51 -0500 schrieb Josh Boyer:
>> >> Why would they leave?
>> >
>> >Because they get little or no support as outlined recently in
>> >
>> I'll address some of your points.  Lots of them are resource issues overall.
>> Leaving because Fedora doesn't have the resources to accomodate a new Spin/
>> effort at the same level as existing ones might very well happen, this is
>> true.  However, there isn't much we can do about that other than to state
>> clearly why we can't do something and ask those wanting it done to pitch in
>> and help.
>Me and a lot of ambitious people are already working hard on the spins.

Yes, we do!

>I don't see what more we can do. If you take a look at my examples, you
>will see there is nothing where we could just pith in and do it. Instead
>these are ether resource issues and we are not to decide on them or
>other things outside our scope.

Not sure what you meant by your last sentence.

>>  * The spins get little support by rel-eng: I had to withdraw the
>>    LXDE spin due to a critical error one day after it was released.
>>    I fixed the problem within 2 days, nevertheless it took 4 weeks
>>    until the new images were created.  rel-eng also denied to test
>>    the spins (although this workflow is written down in the wiki),
>>    but never asked the spins SIG to test them ether. 
>> Rel-Eng consists of about 4 people that are actively working various areas
>> of rel-eng for Fedora.  The one person that creates the spins also happens
>> to be responsible for creating the other 9 spins as well as the two primary
>> spins, the DVD, the CD images, staging it to the mirrors, etc.  One person
>> only goes so far.
>> Testing wouldn't be done by rel-eng, but QA.  QA is similarly understaffed.
>The spins SIG offered to take over testing of the spins, but so far
>there has been no agreement between spins SIG and rel-eng.

Communicate again?  I don't see how/why that would be a problem.

>> Further, you are the primary creator of the Spin.  Why _wouldn't_ you test
>> it?  
>Because me as the primary creator was not even aware of the images.
>Nobody in rel-eng bothered to contact the spin owners after they created
>the images, so the owners were not able to test them.

Did you compose your own for testing?

>>  * Little support by infrastructure: We still have no direct
>>    downloads of the spins, only torrents.  The leader of Fedora
>>    infrastructure even claimed that the "spins are a detriment to
>>    Fedora". 
>> We try very hard to keep the mirror content under 1TB.  We actually have
>> gotten rather close a number of times.  Hosting more content for spins
>> would push us past that.
>I doubt that 2,4 GB which (total sum of Xfce and LXDE for i686 and
>x86_64) make a large difference. And I doubt we couldn't save these 2,4
>GB somewhere if really necessary.

It's not 2,4GB.  If it was, that would imply that your spin is the only
one worth having direct downloads.  It's:

<spin size> * <number of arches> * <number of spins>

It explodes rather quickly when you get into the "why don't the XYZ,
ABC, 8UR spins have direct downloads on the mirrors?" type issues.

Also, it's _really_ hard to drop 2,4GB on the mirrors when you have
a rapidly growing package repository, up to 3 fully supported releases
plus updates, and now throw NFR into the mix.  There is no magical
amount to just "save".

>Right, this was already on another topic, the commitment from Red Hat in
>general. Let's skip this, it doesn't belong here. I have named 6 groups
>in Fedora or 6 issues where there is to little support for the spins.
>You only commented on 2 of the issues.

I can't comment for Marketing, the Design team, or FESCo, so I won't.
(I only counted 5 as well, so I'm missing the 6th?)

>> All of the above seem very much problems with resource contraints.  If
>> we had infinite resources, we could provide equal support for whatever
>> Spin was created.  We don't have infinite resources.  I will go so far
>> as to say that the resources we have are insufficient for our current
>> demand as it is.  Rel-Eng, QA, Infrastructure, Design team, all of them
>> need more people.
>Right, we don't have infinite resources, so we need to arrange them
>better and more suitable.

More suitable for whom?  Just for XFCE?  Just for localized Spins?  How do
you decide these kind of issues?  Do you have some kind of formula you
would like to propose that would make all contention go away or improve it
even a little for a large portion of our user/contributor base?

I'm trying very hard to have an actual discussion about issues, but all
I'm getting out of you is "We need to do better" and "it can't be that
hard" and "ME TOO".


More information about the advisory-board mailing list