** Otavio Salvador :: > >>>>> "humberto" == Humberto Massa Guimaraes > <[EMAIL PROTECTED]> writes: > > humberto> IMHO, there is a series of (serious) problems in such a > humberto> plan, such as: > > humberto> * testing and unstable are not installable by > humberto> non-tech-folk, all the time, really. There can be times > humberto> where they are, but there are some times they are > humberto> not. They break. > > unstable really break sometimes but testing exist to be a always > working version. This is why sometimes things have a so long delay > to enters testing while it has something broken or with RC issues.
Not really. I explain: when a bug goes from unstable to testing (and they do) and renders stuff uninstallable on testing, there is a longer delay where things *will* be broken there, until the corrected version goes there from unstable. > > humberto> * we should not really multiply (space, time, bandwidth) > humberto> needed for our mirrors; right now, some archs are > humberto> endangered because of such hefty requirements. > > we currently have support for partial mirroring using a lot of > packaged tools like debmirror, rsync, mirror and > debpartial-mirror. > > humberto> * we *do* have, after all, "tasks" to install desktops > humberto> and (some, specialized?) servers, without having to > humberto> resort to creating another 30G of repositories. > > I didn't understand what you mean here. Please explain. The problem with Wiktor's proposal is not only mirroring, but storing, building, and transferring (HD, CPU, bandwidth) *separately* what he calls desktop-testing, desktop-stable, etc etc. > > humberto> * finally, the infrastructure necessary to do what you > humberto> ask for is really a job better done by specialized > humberto> derived distros (such as LinEx, Ubuntu, even Ian's own > humberto> Progeny) > > Well yes and no. If we had a place to move the improvements we > need on derivative distributions could be better since we have the > possibility to merge more code and more effort and start to have > more cooperation. > > Debian have all needed structure done. DAK support it very well > and what is really need is only decide what is the rules for > packages migrate to that releases from unstable. No, after you decide that packages can migrate there is a lot of things you should provide: more storage for the now-frozen package, CPU to rebuild that specific version with the specific dependencies in a point in time, bandwidth to transfer it back and forth and to the mirrors. All this without giving anything more than desktop-task. > > I'm not sure if this is good or bad for Debian but is possible to > have it working without so much effort. We will have to agree in disagreeing. -- HTH, Respectfully, Massa -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]