Ext4 is good news
|
| Author | Content |
|---|---|
| sbergman27 Jun 30, 2006 10:18 AM EDT |
This news is great. I like the "new" kernel 2.6 development process. (Use a vendor kernel! Use a vendor kernel!) It has worked remarkably well. It gets new features into the mainline kernel with minimal latency. (Use a vendor kernel! Use a vendor kernel!) And it is quite flexible for the packagers. Conservative, enterprise style distros can pick a kernel, patch it as desired, and standardize on it at any time. The bleeding edge distros can stay as close to mainline as they want to. The down side is that the 2.6 development process actually stifles certain types of development which are more radical, and affect sensitive systems. Forks of individual subsystems are now essential to avoiding the case where major, long term progress is impeded. (Add feature X to ext3? Are you crazy? That would be too disruptive to our existing user base!) For some time, I have been wanting to tell the ext3 developers to just "fork off". In a good way, of course. Did I mention that using a vendor kernel is a good idea? |
| jdixon Jun 30, 2006 10:42 AM EDT |
> Did I mention that using a vendor kernel is a good idea? What do you recommend when your vendor's kernel IS the default kernel (i.e., Slackware)? |
| sbergman27 Jun 30, 2006 11:08 AM EDT |
I guess Patrick has confidence in the one he picked, then. I don't use Slackware for the very reason that it feels like a one man show. To me, the 2.6 development model more or less requires that the final QA be moved to the vendor. This makes a lot of sense to me for a number of reasons. It allows the developers to do what they do best: develop. It moves the responsibility of QA to the ones actually packaging and distributing to "customers". It makes sense to me that the end responsibility should be theirs. The packager/distributors typically have far better facilities for doing QA, as well. Plus they are testing a single, stable set of packages, all together. The patches resulting from the QA should, of course, be sent back upstream, where appropriate, for evaluation, and possible inclusion in mainline. It seems so much more *efficient* to divide the responsibilities that way. I have long felt that those who insist that the vanilla kernel be of commercial quality are wanting Linux to go back to an older, less efficient development methodology. Why waste developers' talent doing grunt work that they don't have the proper resources to do in the first place? |
| jdixon Jun 30, 2006 1:00 PM EDT |
> I don't use Slackware for the very reason that it feels like a one man show. Yes, it is, but I'd say Slackware has demonstrated Patrick's ability over the years. It's not the be-all and end-all distribution, but it is what Patrick wants it to be; stable, fast, and flexible. > To me, the 2.6 development model more or less requires that the final QA be moved to the vendor. Which may be why the default kernel for Slackware is still 2.4. 2.6 has been in testing for a long time, but Patrick has never made it the default. That will probably happen with 11.0 though. |
| sbergman27 Jun 30, 2006 1:55 PM EDT |
Indeed. Considering that he works pretty much alone, his accomplishments are notable. Slackware '96 was my first Linux distro. I was unaware that he was still using 2.4. Makes sense, though. I should probably try out Slackware again to see how it's doing. It could use a bit of a PR makeover, though. (Then again, boring can be good in an OS.) Perhaps if he renamed it to Slackbuntu? IIRC, Slackbuntu is an ancient Minnesotan word that means "Gzipped tar archives are cool". |
| jdixon Jun 30, 2006 5:45 PM EDT |
> IIRC, Slackbuntu is an ancient Minnesotan word that means "Gzipped tar archives are cool". Unfortunately, it's almost 3 months to the day too late to post that to the Slackware lists. :) |
Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]
Becoming a member of LXer is easy and free. Join Us!