Messages in this thread | | | Date | Tue, 22 Jun 1999 21:19:38 -0400 (EDT) | From | "Mike A. Harris" <> | Subject | Re: Linux versioning scheme |
| |
On 22 Jun 1999, Ramana Juvvadi wrote:
>> Coudln't the "incomplete or experiemental" sections of 2.2.x kernels >> already be considered Beta? I mean, I think the beta channel already >> exists in the current versioning system. >> >> brian >> > >That is good for identifying which parts of the linux kernel are beta >( or alpha quality). I am talking about a stamp of approval >from the kernel developers for the overall kernel. Right >now there are only 2 labels -- Development and Stable. I am >taking about refining it a bit more-- development, beta and >stable.
Once again. There _IS_ a beta. Download the 'pre' kernels. You can expand the 'pre' to mean 'prerelease beta of a testing kernel which will be released as stable if there are no major problems reported in x amount of time'.
>At least with the 2.2 series, I think distribution makers >(Redhat, Suse, and possibly others) jumped in too early. At >the risk of sounding bureaucratic, let me suggest a scheme. >I think we should wait for x days ( a week, 10 days pick >your choice) before a beta version is declared stable.
You don't follow Linux development very closely then. Kernels are released in pre's until it is considered stable. Nobody can FORCE people to test the 'pre' kernels though, and many people WONT do so. As such they don't get the exposure that stable kernels do. Due to that, sometimes after a 'pre' kernel has been out long enough, it get's promoted to stable - as you say, only to have 50000 times as many people download it than downloaded the 'pre' beta kernel. Then, due to the larger amount of testing, new bugs are found, some major. I can't see how it could get much better, other than setting up some big warehouse with 1 of every piece of hardware, and forcing Linus to test every possible kernel compilation combination out on every piece of hardware.
It isn't possible, or practical.
>Of course, you can argue that users can a set a rule for >themselves. It just make the life of users a little easier >if the software itself gives more information about its quality.
Even a buggy kernel, or a 2.1.x kernel with values of X > 85 is much less buggy than Windows 95 per se. As such, it is very high quality.
Most major problems are due to weird hardware combinations as I see it from observing the list. As such, there are not a zillion test machines running every single kernel release. People can't afford to run a pre kernel on their 24/7 mission critical hardware if there is a possibility of breakage. As such, the pre doesn't get tested on their machine, so the final release that they download instead, may in fact not work on their hardware combination.
There is just no way to change this, and I don't see anything that is wrong with the current way things are done.
-- Mike A. Harris Linux advocate GNU advocate Computer Consultant Open Source advocate
Tea, Earl Grey, Hot...
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |