From Fedora Project Wiki

< FWN‎ | Beats

(add reference placeholders)
Line 83: Line 83:
[[User:Toshio|Toshio Kuratomi]] completely declined to play and asked: "I'm hereby giving notice that I don't have time to read obvious flamefests anymore.  Once this thread concludes, please summarize whatever the pros and cons are and send it to the packaging committee to discuss and vote on."
[[User:Toshio|Toshio Kuratomi]] completely declined to play and asked: "I'm hereby giving notice that I don't have time to read obvious flamefests anymore.  Once this thread concludes, please summarize whatever the pros and cons are and send it to the packaging committee to discuss and vote on."

Revision as of 01:55, 20 April 2009


In this section the people, personalities and debates on the @fedora-devel mailing list are summarized.

Contributing Writer: Oisin Feeley

Frozen for Fedora 11. Some Packages Still Not Built dist-f11

Jesse Keating announced[1] that henceforth all F-11/ builds would go to dist-f11-updates-candidate and builds from devel/ would go to dist-f12. He asked for concerned parties to check that builds were being properly tagged.

In response to Mike Chambers' question Jesse confirmed[2] that the nightly rawhide composes would consist of Fedora 11 content until the GOLD packages were on their way out to the mirrors at which point the nightly rawhide composes would contain Fedora 12 content.

On a related note Bill Nottingham asked[3] maintainers of a list of packages not yet rebuilt in dist-f11 (with the attendant compiler and strong RPM hashes) to fix them if possible. Jesse Keating provided[4] a slightly more aggressive list as an addendum.

Xorg Hacking Solves DontZap

Peter Hutterer made some valuable contributions to resolving the furore over the disabling of the zapping of the Xorg server via the Ctrl-Alt-Backspace key combination[1].

Tom Callaway drew attention[2] to a blog entry of Peter's which mentioned upstream patches by Julien Cristau (of Debian) to xkeyboard-config and Peter's own patch[3] to Xserver which together make it possible to disallow zapping by default and also to turn zapping on with a

'setxkbmap -option terminate:ctrl_alt_bksp'

. The net result is that it is possible to get zapping to work but the XKB[4] configuration needs to be set up properly and the DontZap option left disabled (as per the new default).

In discussion with Kevin Kofler Peter clarified[5] the situation in which the new settings would take effect. Kevin responded[6] that it appeared that for KDE users zapping with Ctrl-Alt-BkSp would remain as before.

Later Peter answered[7] some questions from Suren Karapetyan about the ability to kill broken X grabs with details about how zapping works.

The above summary of an elegant technical solution ignores the long, and at times vitriolic, complaints about this change. A common trope occurring in some recent threads seems to be that changes are made by Red Hat employees who are implementing changes without community consultation and all work to a common game plan. Seth Vidal challenged[8] the latter assumption:"In a survey of 10 RH employees you will find between 10 and 40 different opinions. sometimes more if you don't ask some of them to confine their comments to a limited amount of time." In any event it's worth noting that the resolution (which filters the "Terminate_Server" action in a manner consistent[9] with the handling of other actions in xkb rulesets) was contributed upstream by a Red Hat employee. As a point of information Kevin Fenzi also made it clear that the change had not been instigated by FESCo.

Minesweeper Certified Solitaire Professionals Satisfied with DVD

Jesse Keating requested[1] help in selecting which packages should be dropped from the DVD image. He suggested some java development packages and games.

Feedback suggested that retaining the games was[2] preferred and dropping the development libraries made sense as the latest versions would be needed[3] and could be obtained from the repositories anyway. Jesse later posted[4] this was sufficient to achieve the desired image size.

A side-issue discussed[5] was the unwieldiness of jigdo as a download method. Callum Lerwick suggested[6] that jigdo would benefit from a userspace ISO implementation.

Presto and DeltaRPM Status

The ability to download binary diffs of RPM packages has been offered[1] for some time now on Fedora through the Presto[2] project and presto-enabled repositories. Interest is high enough in Presto's bandwidth-saving abilities that no fewer than three separate threads were started to ensure that it would function properly for Fedora 11.

Warren Togami asked[3] if Presto would be enabled by default for Fedora 11. Last month (2009-03-21) Jonathan Dieter reported[4] that the use of SHA-256 in rpm had broken deltarpm but that a patched version was available in rawhide. See FWN#166[5] for earlier coverage of the challenges and changes resulting from the introduction of stronger hashes[6]. Jonathan also reported that the changes necessary in infrastructure to build deltarpms had been done. These changes were made fairly rapidly thanks to work done[7] Michael Schroeder, the upstream deltarpm developer. One issue that concerned[8] Axel Thimm was the security with which checksums of deltarpms were being made. Till Maas and Jonathan Dieter provided[9] reassurance that all deltarpms are generated from original rpms which needed to pass all verifications which yum and rpm enforce.

Martin Sourada was excited[10] not just about Presto but also about the slick new PackageKit in Fedora 11. Martin was concerned about the issue of PackageKit and Presto apparently not working well together. A bugzilla entry revealed[11] that PackageKit developer [[User:|Richard Hughes]] quickly created a patch which Martin reported as working.

On 2009-04-16 Bill Nottingham added to the "Rawhide Report" that "[...] rawhide is composed with deltarpms against the prior rawhide. Due to a bug, this is only currently working on i386; it should be fixed for other arches tomorrow. Please test and report any issues."

A Fedora Test Day centering around Presto was also announced[12] by [[User:|James Laska]]. The usual excellent wikipage[13] suggests that Presto can deliver significant bandwidth savings.

Browser Plugins May Strip SELinux Protections

Daniel Walsh asked[1] why mozplugger[2] was being installed by default. He cautioned that mozplugger broke nsplugin and thus SELinux functionality.

An answer posted[3] by Bill Nottingham pointed out the java plugin as the dependent.

Dan worried that while "[a] confined nsplugin is a nice feature for confining plugins downloaded from the network. But if you run openoffice and evince from within nsplugin they get confined, causing the apps to not work properly." In response to Simo Sorce Dan explained that any attempt to write transition rules to enable said applications to work properly would create an easy avenue of attack. Simo wondered[4] if it would be possible to either write a security wrapper to restrict the command line, or to get application developers to honor SELinux labels in some way.

Warren Togami shared[5] that removing mozplugger was "[...] something I always do. It seems to cause more problems than it solves [...]" and James Morris expanded[6] upon this with instructions "[...] on both removing mozplugger and restoring the security protections of SELinux. Simply removing the package isn't enough[.]" James questioned "[...] how a package which breaks a security feature not only made it into the repo, but how it became enabled by default[?]"

A similar issue was raised[7] by Bruno Wolff III about the re-enabling of disabled Firefox plugins. Comments by Martin Stransky suggest this is a feature of mozilla-plugin-config.

Getting Rid of /usr for Fedora 12 ?

Lennart Poettering cheerfully invited[1] any inclined parties to a flamefest over the elimination of the /usr directory. Lennart suggested that recent history indicated that more files were being moved from /usr to / and that confusion between the two was a source of error from some packages.

Enthusiasm for both the flamewar and the proposal was low.

A forceful and well-argued objection was made[2][3] by Konstantin Ryabitsev on the basis that he liked to keep /boot and /usr on their own partitions and use a LUKS-encrypted LVM for everything else. Konstantin emphasized this was especially well-suited to portable machines which need to conserve power and are more likely to need encryption.

Ralf Corsepius invoked[4] the FHS[5] on /usr and the need to contain[6] non-essential packages unavailable at certain boot stages therein. Chris Adams added[7] that symlinking /usr to / had been shown to break rpm.

Lennart explained[8] how /etc could be made read-only and adduced[9] OpenSUSE, Debian and Gentoo as further evidence that a read-only root could be attained. Callum Lerwick pined[10] for the days of floppy disks.

Toshio Kuratomi completely declined to play and asked: "I'm hereby giving notice that I don't have time to read obvious flamefests anymore. Once this thread concludes, please summarize whatever the pros and cons are and send it to the packaging committee to discuss and vote on."