Poll: What is your opinion about the migration to Moblin/RPM
Poll Options
What is your opinion about the migration to Moblin/RPM

Reply
Thread Tools
Posts: 3,319 | Thanked: 5,610 times | Joined on Aug 2008 @ Finland
#81
Now, OBS is certainly something more reasonable than the community/license talk, but doesn't OBS do .debs lately ?
__________________
Blogging about mobile linux - The Penguin Moves!
Maintainer of PyQt (see introduction and docs), AppWatch, QuickBrownFox, etc
 

The Following 3 Users Say Thank You to attila77 For This Useful Post:
Posts: 2,829 | Thanked: 1,459 times | Joined on Dec 2009 @ Finland
#82
Someone has made quite good arguments regarding RPM.

https://bugs.maemo.org/show_bug.cgi?id=9067

Do not talk about it in bugzilla, but in here please. I have no adequate skills or knowledge to speak about this, but Bartosz Taudul seems to have collected his arguments in quite good list.
 

The Following 5 Users Say Thank You to slender For This Useful Post:
Posts: 341 | Thanked: 64 times | Joined on May 2009
#83
voted - I don't care.

i want an open linux computing platform, i will get one either way.

i am an opensuse user anyway, so rpm is no barrier, nor too is the code similarity to fedora rather than debian.

my only concern is that n900 community development is not left to wither on the vine because of the radical differences between freemantle and meego.
 

The Following 3 Users Say Thank You to REMFwhoopitydo For This Useful Post:
smoku's Avatar
Posts: 1,716 | Thanked: 3,007 times | Joined on Dec 2009 @ Warsaw, Poland
#84
Originally Posted by REMFwhoopitydo View Post
my only concern is that n900 community development is not left to wither on the vine because of the radical differences between freemantle and meego.
One radical difference is a switch from GTK+ to Qt... this will cause a huge drop in developer headcount.
Switch from Debian to Fedora (it's not just a package format - it's the whole surrounding ecosystem) is another radical change causing a developers drop.

Will the developers coming from Qt and Fedora domains equalize this?
__________________
smoku @xiaoka.com (SMTP/XMPP) ...:.:....:... pebbled . Poky Fish : sixaxis . psx4m . uae4all
Jolla Phone post-mortem . . . . . . . . . . -> 1+1 VGN-UX390N
 
Posts: 341 | Thanked: 64 times | Joined on May 2009
#85
Originally Posted by smoku View Post
One radical difference is a switch from GTK+ to Qt... this will cause a huge drop in developer headcount.
Switch from Debian to Fedora (it's not just a package format - it's the whole surrounding ecosystem) is another radical change causing a developers drop.

Will the developers coming from Qt and Fedora domains equalize this?
i am delighted about the switch to QT, because i believe it is a much better strategy for the future. this is a change that must be made, and will bring a great number of benefits.
 

The Following User Says Thank You to REMFwhoopitydo For This Useful Post:
Posts: 3,319 | Thanked: 5,610 times | Joined on Aug 2008 @ Finland
#86
Originally Posted by slender View Post
Someone has made quite good arguments regarding RPM.

https://bugs.maemo.org/show_bug.cgi?id=9067

Do not talk about it in bugzilla, but in here please. I have no adequate skills or knowledge to speak about this, but Bartosz Taudul seems to have collected his arguments in quite good list.
I'm no RPM guru, but I'm sorry to say that almost all of those arguments are the results of the author not being familiar enough with present day DEB/DPKG and not the advantage of RPM.
__________________
Blogging about mobile linux - The Penguin Moves!
Maintainer of PyQt (see introduction and docs), AppWatch, QuickBrownFox, etc
 

The Following 3 Users Say Thank You to attila77 For This Useful Post:
rm42's Avatar
Posts: 963 | Thanked: 626 times | Joined on Sep 2009 @ Connecticut, USA
#87
Originally Posted by attila77 View Post
I'm no RPM guru, but I'm sorry to say that almost all of those arguments are the results of the author not being familiar enough with present day DEB/DPKG and not the advantage of RPM.
Can you elaborate? I would be interested in seeing the points he makes rebutted.
__________________
-- Worse than not knowing is not wanting to know! --

http://temporaryland.wordpress.com/
 
Posts: 3,319 | Thanked: 5,610 times | Joined on Aug 2008 @ Finland
#88
The bottom line is things are done differently, so the actual preference is HOW you do something, not WHAT you are doing (that's why the whole DEB/RPM dispute is bogus).

1. I think apt-get dist-upgrade does exactly what he describes. In additions there is an (admittedly seldom used) selection mechanism using which you can specify exact packages to operate on, see dpkg --set-selections (and it's friends, get a and clear) plus apt-get dselect-upgrade

2. This is a default policy. rm -rf / will not tell you you're likely killing yourself either. You have --dry-run if you have doubts what the result will be.

3. sounds like --unpack

4. He doesn't like the debian/debhelper docs. If debian docs are too dry, Ubuntu has very nice packaging tutorials/guides which lack the 'BS' he disapproves of.

5. This has nothing to do with deb. It's just the way debhelpers work - preferring a multi-file setup to a singlefile one, I don't see how this is an objective technical disadvantage. You could unify the debian dir into a single file and build debs from that, nothing deb specific there.

6. dpkg-statoverride (or dh_suidregister). No idea what he's talking about when says it's difficult to make multi-binary releases from one source, all debhelpers do/handle this by default.

7. 3rd party apps. No bearing on deb, anyone can make a superfast deb-based thing too. H-A-M being slow because of all the apt-wrapping is Nokia's choice of not implementing any extra layer of indexing/caching (which is exactly what the 3rd party applications do).

8. debhelpers know about --prefix and dpkg itself has --instdir, it was (again) Nokia's choice to use special /opt handling instead.

That's it in short, off the top of my head. To reiterate, I don't really care either way, just that preferences of doing someting one way does not mean another way of doing the same thing is wrong.
__________________
Blogging about mobile linux - The Penguin Moves!
Maintainer of PyQt (see introduction and docs), AppWatch, QuickBrownFox, etc
 

The Following 6 Users Say Thank You to attila77 For This Useful Post:
Posts: 3,428 | Thanked: 2,856 times | Joined on Jul 2008
#89
I can address a couple I believe:
1. Upgrading packages
There's no notion of "upgrade" action in dpkg/apt. For example, if there are 10
libqt-* packages and "A" is a subset of 5 packages that I have installed and
"B" is a subset of 5 packages that I don't have installed and don't want to
have installed, I cannot easily upgrade just the libqt-A set. I need to specify
each package from the "A" set manually when calling "dpkg -i" or "apt-get
install". RPM has -F option (freshen) that updates only the packages that are
already installed. Poldek (apt-get alike) utility also has "upgrade" command. I
can do "rpm -F libqt-*" and have only libqt-A set upgraded, leaving libqt-B not
installed.
This is somewhat confusing to me. But I believe shows an utter lack of understand of apt (from the man page):
upgrade
upgrade is used to install the newest versions of all packages currently installed on the system from the sources enumerated in /etc/apt/sources.list. Packages currently installed with new versions available are retrieved and upgraded; under no circumstances are currently installed packages removed, or packages not already installed retrieved and installed. New versions of currently installed packages that cannot be upgraded without changing the install status of another package will be left at their current version. An update must be performed first so that apt-get knows that new versions of packages are available.
Edit: Oh, I see - he only wants to do a "apt-get upgrade" of certain packages. I guess I've never run into this problem.

2. Dependency tracking
One can successfully "dpkg -i" a package with an unmet dependency. Such package
will be installed but not "configured" rendering it unusable. RPM won't allow
such action, as it will fail saying there are unmet dependencies. User has a
clear message that there was an error and the package is not left in some state
of limbo. Both apt-get and poldek utilities allow automatic installation of
dependencies though.
This is actually a matter of preference. Having dpkg behave in this way allows me to then do "apt-get -f install" to automatically download and grab the dependencies for that package, and then properly install all of them. No dependency hell (meaning tracking them down myself, or yum installing them one by one), so to speak. What is the equivalent of this in rpm?

His #3 about scripts is an interesting perspective. But it can be both good and bad. This is entirely based on the packager and whether he did a good job or not: But a *lot* of proper uninstallation can be put into prerm scripts and thus, if you don't run them, would leave your system in a very flakey state. If there's an error in the script, then it's obviously a problem, but skipping the script altogether may not be the best answer (although the option *should* probably be there..) because there is no telling what the install scripts could have modified that the remove scripts should be fixing without going into the script and reading it.

As far as his Build documentation (#4), that's entirely perspective. I have never had a problem finding the necessary how-to's to building a deb file properly or the proper formatting of the files.

The Build files (#5), again, never had a problem with this. And the nice thing about the debian/rules file is you can manually edit the exact steps needed to compile a piece of software: in case you get something that is written using a strange system other than cmake/qmake/automake/etc. I personally like having separate files, it's cleaner and easier to read, than a single, long, file to accomplish the same thing.

Somebody else will need to address 6.. I don't understand how RPM can magically allow a user to suid root a binary that debian can't. This looks like something that should be taken care of in a postinst type script and can only be done as root... and the .deb takes and packages up the files with the permissions that they were given when you build the deb package - so if you suid the binary as root, build the deb, and then install it - it will come out suid. I don't understand what he's getting at.

#7 - Speed. Um... ok? There's helpers for RPM.. and there's helpers for Debian... so... who cares?

#8 - Tries to simplify a complex problem. You can't just dump your binary into /opt/whatever-you-want because /opt is not in the $PATH of the users. So, in addition to redefining the prefix to install to (which, by the way, is perfectly possible by editing the debian/rules file to pass whatever prefix options you want), you also need to symlink the proper binaries to the rootfs. Until and unless there is an official and standardized directory structure to place executable binaries: such as /opt/bin and not /opt/package_name or /opt/usr/bin or /opt/MyMommyThinksImSpeshul... etc. Currently I haven't found anywhere that says "This is the exact opt structure your app should look like" - so what we end up with is 30 different applications optified to 30 different sub directories and they can't all be added to the users $PATH unless every application is written to modify that variable every time. I don't see any benefit to using RPM's __prefix over just add --prefix or CMAKE's prefix commands to debian/rules.

#9 - bells and whistles: Yay? I'm sure dpkg/apt have a few RPM don't as well.. what's important here is the basic structure of package installation and dependency management.

At this point.. in those specific developments.. it doesn't matter if you have DPKG or RPM.

The main thing that matters, as has been repeated many, many, times - is the actual structure of the filesystem and OS. I personally have never liked SuSE, Mandrake's, or Red Hat's structure. I also have had no end of trouble using Yast or URPMI and up2date.

However, I've never had any complaints with YUM. So I don't care if either YUM or APT-RPM are used in the new system - they both work fine for me. As long as the OS structure makes sense.
__________________
If I've helped you or you use any of my packages feel free to help me out.
-----------------------------------------------------------------------------------
Maintaining:
pyRadio - Pandora Radio on your N900, N810 or N800!

Last edited by fatalsaint; 2010-02-17 at 15:53.
 

The Following User Says Thank You to fatalsaint For This Useful Post:
SubCore's Avatar
Posts: 850 | Thanked: 626 times | Joined on Sep 2009 @ Vienna, Austria
#90
Originally Posted by rm42 View Post
Can you elaborate? I would be interested in seeing the points he makes rebutted.

the few things i noticed:

1. Upgrading packages
There's no notion of "upgrade" action in dpkg/apt.
apt-get upgrade maybe??
(at that point i already knew he doesn't really have much experience with debian packaging...)


2. Dependency tracking
One can successfully "dpkg -i" a package with an unmet dependency. Such package
will be installed but not "configured" rendering it unusable. RPM won't allow
such action, as it will fail saying there are unmet dependencies.
dpkg -i won't install stuff without dependencies. you have to --force-things or --ignore-depends for that to happen.
dpkg -i without either of those additional parameters will fail when encountering unmet dependencies.


3. The scripts problem.
Sometimes a package with broken pre-uninstall script is installed.
dpkg --force-remove-reinstreq


4. Build process documentation [...]
On the other hand, all the deb package creation documentation heavily orbits
around the debian policies about freedom, distributions, maintainers, and other
uninteresting BS.
mhm...
centralized debian documentation > fragmented rpm documentation IMO


5. The build files [...]
now that part's a rant from someone in unfamiliar territory, nothing more.


7. Speed
that's a maemo-specific problem. actually, it's an application-manager specific problem. has nothing to do with apt/dpkg helpers. (apt-get install something in xterm will NOT trigger a DB rebuild, it IS cached).


8. The /opt problem
i doubt that just using .rpm and prefices/macros will magically solve this problem. many things won't "just work". that's hopelessly optimistic
__________________
"What we perceive is not nature itself, but nature exposed to our method of questioning."
-- Werner Karl Heisenberg
 

The Following User Says Thank You to SubCore For This Useful Post:
Reply

Tags
rabble-rousing, rpm vs. deb war, rpmligion vs debligion, vote attila77


 
Forum Jump


All times are GMT. The time now is 09:27.