Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
Alle 13:50, sabato 24 dicembre 2005, Peter ha scritto: Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. As a user, I wouldn't normally write to this list, but you asked for our opinions. Apart from the fact that I upgrade the kernel more often than I upgrade nvidia-kernel, and as many have pointed out this means reistallyng glx, nvidia-settings and nvidia-xconfig even if not needed, there's also another important point: This unified ebuild will lead me to upgrade the package even if it is just a rev-bump for issues related to (e.g.) nvidia-settings, which I don't even use!! Or do you expect that users always look at the Changelog to see if the upgrade really applies to them. How could this improve the user's experience? Some may not agree with the idea of turning it in a meta-ebuild, but if this must really be done, I think that would be the only way to keep everyone satisfied. -- Move -- Proudly abusing Gentoo Base System version 1.12.0_pre12 Linux 2.6.14-gentoo-r5 -- gentoo-dev@gentoo.org mailing list
[gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On Sat, 24 Dec 2005 13:09:36 +0100, Carsten Lohrke wrote: On Saturday 24 December 2005 12:34, Peter wrote: THAT is a very reasonable comment! Not at all. Meta ebuilds are a provisional and fugly workaround as long as we have to wait for proper sets and only to be used for a larger set of packages. Wrapping three or four ebuilds with another one, just for the sake of lazy people having only to emerge a single ebuild is ridiculous. The only valid point in the original idea to merge the nvidia-* packages is to reduce the overall number of packages in the repository. As you heard the voices against it, ranging from none to valid reasons, everything left is to bury the idea and to close the bug. Carsten Would you please add the comments to the bug report? Or, may I copy them? Please advise. Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. If you are against meta ebuilds, what is your opinion on KDE? Instead on 9 (or so) packages, there are now over 250! Are 250 separate ebuilds better? I cannot think of another distro that slices up KDE that way. Meta ebuilds at least allow the user the ability to opt out of trying to decide which ebuilds to emerge! I always used to use CVS to update my KDE source tree, then compile only the changed modules. I could have a whole updated KDE inside an hour. Now that is performance! Here, with the unified nvidia, the intent was to REDUCE ebuilds and simplify installation process. I thought the recommendation of a meta nvidia ebuild is a worthy one worth consideration. IMHO sometimes the desire to fine tune things and optimize things goes a little over the edge. nVidia upstream combines all the products together in their .run files. There is minimal time difference between having the entire suite installed versus each one individually. And, from a user's point of view, what could be simpler? Anyway, I appreciate your feedback. -- gentoo-dev@gentoo.org mailing list
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On Saturday 24 December 2005 13:50, Peter wrote: Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. Considering that we aren't paid, I think that every _affordable_ effort should be made, but making more complex maintainership for devs just to satisfy a couple of users, when the advantages are really minimal, it's not exactly a good choice, IMHO. Here, with the unified nvidia, the intent was to REDUCE ebuilds and simplify installation process. I thought the recommendation of a meta nvidia ebuild is a worthy one worth consideration. nvidia-glx depends on nvidia-kernel already, no? That would be enough, for me. nVidia upstream combines all the products together in their .run files. There is minimal time difference between having the entire suite installed versus each one individually. Well depends how you see it. If you just build it when you update the drivers, yeah there's a minimal difference. But if you have more than one kernel (for whatever reason), and you want to have the latest kernel on all of them, it's way faster to just use nvidia-kernel. Then there's the point I've already said, about mixing the kernel-level with generic userland stuff: for Gentoo/FreeBSD I need it to be split, or I'd have to recreate a copy ebuild especially for FBSD... and that not only sucks from an user POV but also from a maintenance POV. -- Diego Flameeyes Pettenò - http://dev.gentoo.org/~flameeyes/ Gentoo/ALT lead, Gentoo/FreeBSD, Video, AMD64, Sound, PAM, KDE pgptjX1EPKlJN.pgp Description: PGP signature
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On Sat, Dec 24, 2005 at 07:50:51AM -0500, Peter wrote: Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. customer n : someone who pays for goods or services [syn: {client}] When did we start selling Gentoo? (Admittedly we sell optical media via the Gentoo Store, but the software is still free-as-in-beer) -- Jon Portnoy avenj/irc.freenode.net -- gentoo-dev@gentoo.org mailing list
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On Saturday 24 December 2005 13:50, Peter wrote: Would you please add the comments to the bug report? Or, may I copy them? Please advise. Feel free to do so. Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. The only valid issue I read about against a single nvidia-driver package was Diego's regarding FreeBSD. On the other hand having packages split or merged only because it can be done, doesn't make much sense. Regarding happiness, merry x-mas and best wishes to everyone, but I care about maintainability in the first place. You may be interested in GLEP 21¹, a very user friendly approach to easily group user defined packages. If you are against meta ebuilds, what is your opinion on KDE? Instead on 9 (or so) packages, there are now over 250! Are 250 separate ebuilds better? I cannot think of another distro that slices up KDE that way. Meta ebuilds at least allow the user the ability to opt out of trying to decide which ebuilds to emerge! The split was due and having meta packages of some sort was necessary due to the amount of packages. The problems are present though: re-emerging and un-emerging meta packages doesn't affect their child packages. For reasons why the split ebuilds are an advantage please refer to The KDE Split Ebuilds HOWTO². The big downside of (the large number of) split packages is the affect on the code base of the stable Portage versions (and the current layout of the repository), which apparently is not created having scalability issues in mind. I always used to use CVS to update my KDE source tree, then compile only the changed modules. I could have a whole updated KDE inside an hour. Now that is performance! But this has nothing to do with providing a large user base with a reasonable stable set of packages. Here, with the unified nvidia, the intent was to REDUCE ebuilds and simplify installation process. I thought the recommendation of a meta nvidia ebuild is a worthy one worth consideration. As explained above, it isn't. IMHO sometimes the desire to fine tune things and optimize things goes a little over the edge. nVidia upstream combines all the products together in their .run files. There is minimal time difference between having the entire suite installed versus each one individually. And, from a user's point of view, what could be simpler? Anyway, I appreciate your feedback. If Diego could explain what the FreeBSD problem is and if he can resolve it, personally I don't see a valid reason against having a unified nvidia-driver package. I assume all the steps Portage is doing to install those packages (and uninstall previous versions) will take more time than having it bundled in a single ebuild, anyways. But raising the number of packages by a crappy meta ebuild (sorry, lazy users don't count) - no. Carsten [1] http://www.gentoo.org/proj/en/glep/glep-0021.html [2] http://www.gentoo.org/doc/en/kde-split-ebuilds.xml#doc_chap3 pgpjar6Psrvfo.pgp Description: PGP signature
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On 12/24/05, Peter [EMAIL PROTECTED] wrote: On Sat, 24 Dec 2005 13:09:36 +0100, Carsten Lohrke wrote: On Saturday 24 December 2005 12:34, Peter wrote: THAT is a very reasonable comment! Not at all. Meta ebuilds are a provisional and fugly workaround as long as we have to wait for proper sets and only to be used for a larger set of packages. Wrapping three or four ebuilds with another one, just for the sake of lazy people having only to emerge a single ebuild is ridiculous. The only valid point in the original idea to merge the nvidia-* packages is to reduce the overall number of packages in the repository. As you heard the voices against it, ranging from none to valid reasons, everything left is to bury the idea and to close the bug. Carsten snip Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Remember that users are your customers. Every effort should be made to keep them happy. This is a voluntary based distro. Gentoo devs are users too only with added responsability (there's more to it but just for the sake of simplicity ...). I cannot speak for the maintenance side of things though. If you are against meta ebuilds, what is your opinion on KDE? Instead on 9 (or so) packages, there are now over 250! Are 250 separate ebuilds better? I cannot think of another distro that slices up KDE that way. Meta ebuilds at least allow the user the ability to opt out of trying to decide which ebuilds to emerge! Please check this url : http://packages.debian.org/stable/kde/. Actually, Gentoo was one of the very rare distro that was bundling kde-base, kde-network etc all in one big package. I always used to use CVS to update my KDE source tree, then compile only the changed modules. I could have a whole updated KDE inside an hour. Now that is performance! That is the whole point of the split kde ebuilds. It should be even faster than your method when conf-cache is fully implemented. I don't remember if it is finished though. It is also automated, less error prone and intergrated in our favorite package manager. Thank you KDE team for the good work by the way. Here, with the unified nvidia, the intent was to REDUCE ebuilds and simplify installation process. I thought the recommendation of a meta nvidia ebuild is a worthy one worth consideration. It simplifies the installation so to speak but for many gentoo users that changes their kernel often it's a pain. How would your ebuild react to module-rebuild ? IMHO sometimes the desire to fine tune things and optimize things goes a little over the edge. nVidia upstream combines all the products together in their .run files. There is minimal time difference between having the entire suite installed versus each one individually. And, from a user's point of view, what could be simpler? I agree with this but from a user point of view having both would be better. A meta ebuild with correct use flags that pulls the necessary dependency but leaves us with the flexibilty of easy kernel mangling is good even if it's a a provisional and fugly workaround when it's all we have for this type of situation. Albeit, when you remove the meta ebuild it doesn't remove it's dependency except if you run the very scary depclean and for good reasons. The user would be left with all the modules lying around when he though that it was removed. Anyway, I appreciate your feedback. Sure no problem. As a long time Gentoo user, I chose this distro for it's flexibilty and modularity not it's simplicity. Gentoo devs have a habit (policy ?) of following upstream as long as it fits well with the distro. Moreover, bugzilla is not used for discussing proposals. wkr and happy holidays Jean-Francois -- gentoo-dev@gentoo.org mailing list -- gentoo-dev@gentoo.org mailing list
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
On Sat, 24 Dec 2005 07:50:51 -0500 Peter [EMAIL PROTECTED] wrote: | Also, I find it absolutely fascinating that the only people against | this concept are devs, and the only people for it are users. Remember | that users are your customers. Every effort should be made to keep | them happy. Hardly surprising. Our 'customers' don't always know what's best for them, especially when it comes down to low level issues like this one. -- Ciaran McCreesh : Gentoo Developer (I can kill you with my brain) Mail: ciaranm at gentoo.org Web : http://dev.gentoo.org/~ciaranm signature.asc Description: PGP signature
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
Peter wrote: Also, I find it absolutely fascinating that the only people against this concept are devs, and the only people for it are users. Please see Message-ID: [EMAIL PROTECTED] (for those using MUAs that suck, it's Date: Sat, 24 Dec 2005 03:16:53 -0600, From: Dale [EMAIL PROTECTED] in this thread). If you are against meta ebuilds, what is your opinion on KDE? Instead on 9 (or so) packages, there are now over 250! Are 250 separate ebuilds better? I cannot think of another distro that slices up KDE that way. Meta ebuilds at least allow the user the ability to opt out of trying to decide which ebuilds to emerge! You seem to be confused by split ebuilds and meta ebuilds. Cheers, -jkt -- cd /local/pub more beer /dev/mouth signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] Re: Re: Re: Unified nVidia Driver Ebuild ready for testing
I'm just a user, but I personally would prefer the three separate ebuilds. If a meta-ebuild was included as an additional way to build, that'd be fine. I update whenever a new version comes out, but only build -kernel after updating the kernel. This makes sense as being the most efficient way to go. As for the extra packages, I don't use them and don't intend to. I know they may be small, but that's just one more file/build that I don't have to worry about and that clutters the system; I'm very picky about my system. Just my opinion on this. Have a safe holiday break. -- gentoo-dev@gentoo.org mailing list