From Fedora Project Wiki

< FWN‎ | Beats

Revision as of 19:20, 17 May 2009 by Ush (talk | contribs) (fix missing refs)

Developments

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

Contributing Writer: Oisin Feeley

Broken Dependency Brouhaha

The deliberate introduction of a broken dependency by Richard W.M. Jones resulted prolonged discussion and two FESCo discussion items tabled for the 2009-05-15 meeting. One of those items was the possible removal of "provenpackager" status from Richard.

Michael Schwendt noticed[1] that an update for libguestfs[2][3] had been pushed by developer Richard W.M. Jones in the full knowledge that Fedora 10 users would need to import a Fedora 11 qemu package. An anonymous comment on Bodhi situated the decision to release the update as an example of Richard not respecting the release process. Richard argued[4] that as the libguestfs package was completely new only those aware of what they were doing would install it (and consequently would be aware that they needed the qemu from Rawhide or Fedora 11.)

A strong reaction against "[c]reating broken deps when you know they won't be corrected[...]" ensued[5] and led[6] to Seth Vidal deciding to question Richard's suitability as a "provenpackager" on the basis that he lacked common sense.

A sidethread on the advantages of introducing dependency-checking was started by drago01. While Josh Boyer agreed[7] that it would be useful he asked for help in solving the difficult problems which he listed.

The 2009-05-15 FESCo meeting resolved[8] that Toshio Kuratomi and Richard W.M. Jones should draft a Packaging Guideline for approval by the Fedora Packaging Committee. The meeting also decided that as Richard's introduction of a broken dependency was made in the absence of a clear prohibition against such actions, and he was clear that it would not recur, then no sanction should be taken. The handling of similar requests in the future were agreed to be best handled on a case-by-case basis.

Richard added[9] that the necessary back-porting of changes to qemu in Fedora 10 were going to happen. Currently the update has been revoked.

Verilog Emacs Add-Ons

The prime mover behind the Fedora Electronic Lab Spin, Chitlesh Goorah, asked[1] for feedback on splitting-out "verilog-mode" into a separate package so that upstream changes could be tracked more rapidly. This would also have the benefit of laying the groundwork to support OVM and VMM (see FWN#161[2]).

Jonathan Underwood made[3] some good points concerning the danger of missing out on emacs trunk integration of such packages if they were split out. He suggested instead: "[...] a packaging strategy whereby we don't rip out verilog-mode from the core emacs packages, but we can also have an add-on package which contains the latest and greatest verilog-mode which, if installed, is loaded in preference to the one from the core emacs packages[.]" This seemed to be accepted as a positive direction by Chitlesh and a review of the emacs-verilog-mode package was started[4] by Jonathan.

Jerry James raised[5] the issue of XEmacs also having its own version of the package, due to byte-code divergence between Emacs and XEmacs, and also some GPLv2 versus GPLv3 compatibility issues.

Open JDK7 Experimental Package

Lillian Angel asked[1] where the OpenJDK[2] team should post their unstable java-1.7.0-openjdk package: 1)to RPMFusion; 2) to a personal FedoraPeople page; 3) to the main Fedora repositories.

Lillian disliked the last option: "I am not keen on getting this package pushed into Fedora since java-1.6.0-openjdk already exists, and jdk7 will not be stable until sometime after Feb 2010[3]."

Following several suggestions it was decided[4] that a personal FedoraPeople repository was the best solution as there would be six or seven packages with no interdependencies.

Making Noise About Moksha

When Dimi Paun continued[1] to report problems using PulseAudio (see FWN#[2]) responses suggested[3] that his use of non-Free Flash or tweaking of GStreamer settings was responsible. Debugging using gstreamer-properties to ensure that "pulsesink" or "autoaudiosink" was the default sink was recommended[4].

Lennart Poettering wanted a bug filed instead of posts to @fedora-devel and when Dimi explained that Bugzilla was too slow and he had already spent a lot of time on the problem Rahul Sundaram suggested[5] using Bugz instead.

Criticism of the display of possibly thousands of "CLOSED" bugs by Bugz led Tom Callaway to offer[6] the hope that Fedora Community will allow developers to "[...] show new/open packages only on a per package basis[.]" This occasioned[7] some apparent criticism from Rahul Sundaram of a lack of openness "[...] it is a giant silo [...]" around the development of Fedora Community[8]. Tom Callaway offered[9] a list of resources to contradict this. When Rahul returned[10] with the criticism that there "[...]is definitely a big lack of communication on this development with the rest of the Fedora community. There was a very brief mail to fedora-announce list but how much input are you getting input from Fedora maintainers whose job this is supposed to make easier?" there was a distinct lack of enthusiasm for more aggressive marketing. Josh Boyer reaffirmed the involvement of several developers with large package lists and expressed[11] a fear that bike-shedding would result from any more exposure. Paul W. Frields pointed[12] to a useful interview[13] with Luke Macken about the Moksha web-application framework upon which Fedora Community is being built.

Moksha is built on a collection of python-based web-frameworks and uses Orbited instead of AJAX to connect rich web applications to servers. Reportedly this is more responsive than AJAX techniques.

A test instance of Fedora Community and AJAX was reported[14] by Tom Callaway to be up. He emphasized that it was a test instance, currently not to be relied upon at all and a disinclination "[...] to spend time wading through the `OMG THIS IS SLOWER THAN BUGZLILLA!!!1!'" reports.

Be Excellent to Each Other

Regular readers are no doubt aware that flamewars have become more common on @fedora-devel. Project Leader Paul W. Frields posted[1] to the @fedora-advisory-board that the FAB[2] had decided to deal with the "[...] degradation in tone and signal [...]" by appointing moderators.

Mike McGrath worried[3] that this would constitute an extra burden for board members and also objected to any censorship on principal. [User:Mmcloughlin|Mark McLoughlin]] wondered how posters warned privately by moderators that their behavior was problematic could defend themselves. Seth Vidal replied[4] that this was not a court of law and that problems with moderators could be reported to the board. Later posts along these lines drew[5] a response from Luis Villa which argued strongly that over valuing one's own liberty was a problem: "Or to put it another way: The Fedora community exists to work together towards some common goals. Sometimes, in the name of reaching those goals, you have to be polite and adult towards others so that you can work efficiently and constructively with those other people even when you disagree with them, and work with them in the future after you have stopped disagreeing. This use of words like 'freedom' and 'oppression' suggests to me that some people think their highest reason for being here is about them. It's not about you, it's about working together to build something bigger and better than you. And if you can't play nicely with others in the name of those bigger and better things, or don't understand why sometimes you have to play nice in order to get to those bigger and better things, then maybe this isn't the right place for you."

Paul W. Frields reported[6] that a good deal of work led by Kevin Fenzi was going on to moderate the IRC channels. A later post made[7] by Max Spevack referenced IRC bans in the #cobbler channel and suggested that Red Hat employees needed to be tough-minded and hold themselves to higher standards than other contributors.

Best Way to Store Information Across Desktops

Kushal Das requested[1] tips on making a truly cross-desktop application.

Adam Williamson noticed that many applications were storing information in ~/.config files and Mathieu Bridon provided[2] the information that this was an XDG[3] spec from freedesktop.org which resulted in replacing a plethora of .app directories with only two: .config to store configuration and .local/share/ to store data.

Jaroslav Řezník pointed[4] to work by the KWallet and gnome-keyring developers to develop[5] a single-sign-on solution on top of a DBUS-based protocol.