Core Packages in Violation of the Fedora Naming Guidelines

[Fernando Nasser]

Jesse Keating wrote:
> On Wednesday 12 July 2006 15:52, Nicolas Mailhot wrote:
>   
>> If you move it to the version you're breaking jpp -> fc upgrade paths
>> If you put it as Provides it's ok from the package manager POW but not
>> for users (there are reasons why we use long descriptive filenames and
>> not DOS-like 8:3 names)
>>     
>
> In which ways are we breaking it?  jpp has foo-2.6.0-6jpp.  Fedora has 
> foo-2.6.0.6jpp-1.fc6. 
Why?

I've answered your question, you did not answer mine.


>  2.6.0.6 is rpmnewer than 2.6.0-6, upgrade path exists.  
>   
Where does the version ends and the relase begins in this?

6jpp is part of the release, not the version, and it can be prefixed.

What do you do with    foo-2.7-0.cvs20060123.4jpp ?


> If jpp issues 2.6.0-7jpp, its going to be newer than what FC provides yes, 
> but do we want users picking up that package?  Shouldn't they stay with the 
> FC provided one?  Or do you want it replaced and then replaced again when FC 
> bumps the package?
>
>   
We want the upstream package to take precedence, as fixes go there 
first.  Once we import that version and produce a pre-compiled (AOT) 
equivalent, that takes precedence.


> So then put it in the name rather than the version.
>
> "Upstream": foo-2.6.0-6jpp
>
> Name: foo-6jpp
> Version: 2.6.0
> Release: 1%{?dist}
>
> Provides: foo = 2.6.0-6jpp
>
>   
Versions in package names?  The namespace will explode in no time, no 
history, imagine adding new packages all the time to the bug repository 
lists, and so on.  Huge implications.

We do have a few release-marked packages but that is done only as an 
extreme measure due to incompatible API changes and usually for a 
limited time (the versioned package is just kept for backward 
compatibility purposes for the least possible time).


Regards,
Fernando