Topic: Why RPM5?

[robxu9]

Just wondering why Ark is using RPM 5 compared to RPM 4.8.

[Piki]

Being that I'm not a developer, I can't answer in detail. However, I do know that it is easier to customize RPM5 to use LZMA Compression. This makes it easier for us to provide smaller packages for download from our repositpories, and also to fit more on our ISO Images, which offers us a wider range of possibilities when coming up with different releases of our 2010.1 release.

I will point our head developer here when I can contact him -- he is very busy with his real life job, and so often is not around, but when he is, I will ask him for more detail on why we're choosing RPM5.

[robxu9]

Hmm... (A late reply,) but isn't that unreasonable? RPM 4.8 includes xz compression, something even RPM5 does not provide yet. If xz is used on all distributions and replaces lzma, wouldn't you stay with RPM4?

(By the way, I know you have a lot of packages in your dockyard-devel repository, but I'm just pretty uncertain about that. That would mean that all other users wouldn't be able to upgrade from 2008.1 unless you provided an RPM5 package in the dockyard 2008.1 repository...)

[Piki]

I can't exactly answer that one because I'm not familiar with xz. But if xz is better, why didn't bero see it there when he was using 4.8? And why are the packages now smaller?

As for upgrading from 2008.1 to dockyard-devel, that wouldn't be possible anyway because of the new init system (check the Projects tab of this site). It is bero's custom version of init that will have the system booted up within 3-7 seconds as opposed to the 15-60 seconds of the regular init. Since this is a HUGE adjustment for any upgrade, it could corrupt the current install of 2008.1. So, even with RPM5 in the 2008.1 repos, the only way to cleanly upgrade would be to backup all data, download the new ISO from dockyard-devel, and install the new ISO. I wish we could come up with a clean way to install, but I think when users see all the advantages to the new system, they will not mind having to do it this way.

[robxu9]

hm, sounds good the test iso with the faulty installer, is it going to get updated? I'd like to test it

[Piki]

The installer itself is not faulty, it's the same exact installer that we've used in previous releases. The only thing that changes from release to release in the installer (aside from the fact that the current installer supports 64bit) is the kernel; the possibilities are a faulty kernel, a faulty kernel module, or completely incompatible hardware (the last is unlikely as 2008.1 worked on my old computer but froze up on the test ISO). The link to the test ISO will not need to be updated as the name of the file will be the same; when bero uploads the new installer, I will include a link to the 32bit ISO (if he remembers to build one this time around) and either he or I will make an announcement on the Announcements & News forum.

[bero]

Just wondering why Ark is using RPM 5 compared to RPM 4.8.

Hi,
sorry for the time it took to reply, as Piki pointed out, I'm currently 100% busy with my day job.

There's 2 components to the choice of rpm5 -- one is technical: rpm5 always tends to be a bit ahead of the curve in terms of features (e.g. lzma support was in a few months before rpm 4.x started looking at alternatives to gzip and bzip2; xz support is in as well these days, and given lzma and xz use the same compression algorithm, it didn't break any packages) and stability (the changes that have gone into rpm5 HEAD are likely to eliminate rpmdb corruption problems for good once they're fully finished).

The other component is simply trusting the maintainers. Jeff has maintained rpm for the last decade and while there were occasional problems, he's the one who came up with the things that make rpm what it is, and has always done a good job of keeping it up to date with current needs. He also really cares about the project, if you send him a bug report with sufficient detail, you almost always get a reply the same day.

In return for it, his former employer (Red Hat) fired him (mostly over him giving replies deemed unacceptable to users who didn't supply sufficient information with their bug reports and/or who disagreed with some design choices -- but he's a developer, not a tech support person...) and handed rpm 4.x over to someone else. That's how rpm5 started.

Now Panu (the primary rpm 4.x maintainer) is also a good guy and what happened to Jeff is not his fault -- but his commitment is related to his employment, and if he gets fired or reassigned, I doubt rpm 4.x will continue receiving sufficient maintenance. Jeff, on the other hand, has proven that he cares about rpm and keeps working on it even now that there's no job incentive for him.

That's pretty much my view on things - of course there may well be counterarguments. Do you have any valid reasons why we should be using 4.x?

[robxu9]

eh, no, not really. Hey, it's good to see you back on the forums