Time is something I'm prepared to offer, though my black magic skills are limited. I'll check to see if I can make CPU time available, but that brings up another questions: how do we establish trust? What process is reasonable to verify that packages I build, for example, are trustworthy? I have a handful of packages I'd be happy to send over this weekend (MySQL 4.0.18, etc) but how would the project prefer I sign them, what quality control steps do I need to verify first, etc?
I was also thinking of somthing automated, and while manual intervention still would be required, it would hopefully reduce the cycle time significantly. Ideally, CentOS would have a system that would:
1. Automatically detect when an update has been released (via parsing the announcement e-mail from Red Hat or something?) 2. Download the SRPM. 3. Build a first-pass RPM and notify the package maintainer when it is ready for review. 4. Depending on the maintainer's review, the package is either: a. released (another scripted process which automates GPG signing and transferring the RPM to the appropriate repositories and sends out an e-mail to the list about its availability after an appropriate time-lapse so that the mirrors have a chance to mirror it) or b. rejected, pending modifications and another build cycle.
This would at least take some of the load off the maintainer's shoulders and - hopefully - most updates would rebuild cleanly and not need modifications. When we know in advance that package X needs mods Y, we could make the system aware of it.
All we need is someone talented enough to craft such a system. I can admit without too much shame that it's mostly beyond my abilities. Anyone volunteers out there? ;)
Lance Davis wrote:
What can the community do to help improve turn-around time on Redhat errata?
Volunteers are always welcome :)
The reason the kernel takes so long is that it just takes that long to build and test.(likewise openoffice)
If you have some decent hardware then it should be quicker ...
As regards the other upodates it is a question of checking whether oatches apply cleanly, removing trademarks where necessary and testing.
An automatic system is required, buyt at the end of the day manual intervention will probably always be needed.
Lance