Merge lp:~mrand/ubuntu-seeds/platform.lucid into lp:~ubuntu-core-dev/ubuntu-seeds/platform.lucid
| Status: | Rejected |
|---|---|
| Rejected by: | James Westby on 2012-03-08 |
| Proposed branch: | lp:~mrand/ubuntu-seeds/platform.lucid |
| Merge into: | lp:~ubuntu-core-dev/ubuntu-seeds/platform.lucid |
| Diff against target: |
22 lines (+5/-5) 1 file modified
desktop-common (+5/-5) |
| To merge this branch: | bzr merge lp:~mrand/ubuntu-seeds/platform.lucid |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Martin Pitt | 2010-03-31 | Approve on 2010-04-15 | |
| James Westby | 2010-03-31 | Pending | |
|
Review via email:
|
|||
Commit Message
Change cups/foomatic/
| MarcRandolph (mrand) wrote : | # |
| James Westby (james-w) wrote : | # |
Martin, what do you think of this change?
Thanks,
James
| Martin Pitt (pitti) wrote : | # |
Looks okay to me in general, but I'm a bit nervous about it for Lucid. Spooler-less operation isn't recommended by upstream (it can cause lockups pretty easily), and we have not extensively printing from the desktop without a cups server. As a random example, it's conceivable that system-
Can we postpone this to lucid+1?
| Daniel Richard G. (skunk) wrote : | # |
Bear in mind that cups et al. will still be installed for most users. The only ones who may run into potential problems are those who specifically exclude these packages, or installed the metapackage with --no-install-
FWIW, while I haven't been working with (Xubuntu) installs where cups is not present, I've been using it with cups removed from the rc.d startup symlinks, and have never noticed any ill effects on printing or other aspects of the system.
| Martin Pitt (pitti) wrote : | # |
> Bear in mind that cups et al. will still be installed for most users.
Yes, of course. I was think more along the lines of "as long as someone has ubuntu-desktop installed, it should be a supported configuration". And I pointed out that we cannot currently ensure how well this configuration works in various cases.
| Daniel Richard G. (skunk) wrote : | # |
It would have been nice if someone had brought this up in the 16 months the original bug has been posted.
Martin, if you'd like me to take an Ubuntu install, take out CUPS, and check that all the printer functionality continues to work, I can do that.
That aside, "supported configuration" means as much that we are committed to viewing any potential problems with spooler-less operation as bugs, and to fixing them, as opposed to saying "wontfix, please install cups, kthxbye."
For my part, postponing this to a Lucid point release wouldn't be so bad. But I think it would be really dropping the ball if this doesn't make it into an LTS until the next one (in 2012, is it?), when the whole point of the original bug report was to simplify large installations (which favor LTS editions) by centralizing the spooler.
| Martin Pitt (pitti) wrote : | # |
Daniel Richard G. [2010-04-01 16:04 -0000]:
> It would have been nice if someone had brought this up in the 16
> months the original bug has been posted.
True that, but we have to take it as it is now..
> Martin, if you'd like me to take an Ubuntu install, take out CUPS,
> and check that all the printer functionality continues to work, I
> can do that.
That would be great indeed!
> That aside, "supported configuration" means as much that we are
> committed to viewing any potential problems with spooler-less
> operation as bugs, and to fixing them, as opposed to saying
> "wontfix, please install cups, kthxbye."
Of course they would be bugs, but two weeks before final freeze we do
not have time to create more bugs and fix them all. :-) So we need to
be cautious here.
Thanks,
Martin
--
Martin Pitt | http://
Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org)
| Daniel Richard G. (skunk) wrote : | # |
Okay, I set up a Lucid beta1 system with an up-to-date GNOME desktop (minimal install + equivs packages that Provides: cups et al. + apt-get ubuntu-desktop), and observed the following:
General:
* With only cups-{client,bsd} installed, there is no /etc/cups/ directory, which is not so good---in a spooler-less setup, you're going to want to have client.conf and (maybe) lpoptions in there
system-
* Default value in "Connect..." server entry is localhost instead of /var/run/
* I can enter the server name of my centralized CUPS server, connect to it, and see the associated printers
* Server -> Settings tosses up an error dialog:
CUPS server error
There was an error during the CUPS operation: ".
* Trying to tweak other server/printer settings (e.g. printer properties) asks for a CUPS server login, which is as it should be
* Entering the server name into s-c-p allows you to browse the printers available at that server, but the server isn't saved---other apps aren't aware of it, and the next time you start s-c-p, it's back to "localhost" (ideally, s-c-p would create/modify /etc/cups/
* Trying to set the system default printer asks me for a CUPS login, which is incorrect; all this should need to do is create/modify /etc/cups/lpoptions (I suspect s-c-p is presuming "system" to refer to the CUPS server, not the local system, an assumption that breaks in spooler-less operation)
* Trying to set the "personal default" printer goes through, but I don't get the star over the printer selected, and apps don't recognize it
* The "Manage Print Jobs" dialog seems to work, although because the CUPS server where I work is a frontend for an LPR queue, I wasn't really able to watch job entries pop up
Once I created /etc/cups/
* I can select a "personal default" printer (setting the system default still asks for a CUPS login)
* Applications are aware of the printers at the remote server, as well as my default choice
* All the various application "Print Setup" options work (tried Firefox and OO.org)
* Printing works!
All in all, there's a couple wrinkles, but nothing showstopping. The s-c-p man page even states "As a result, it [system-
It would be nice to be able to configure the central spooler using s-c-p, and I can certainly file bugs on this and all the other nits I found. But sites using a central spooler would set this up in client.conf anyway, so it wouldn't really matter. (And FWIW, even then, remote spoolers are already supported by the GUI better than I had expected.)
| Martin Pitt (pitti) wrote : | # |
Thanks for testing!
So, it seems to work well enough, but still leads to some confusion.
I still wouldn't like to rush it into lucid, a week before final freeze. For local customization you can just remove /etc/rc2.d/S50cups, which is not that much harder than removing the packages (and you need to know what you are doing anyway, with having to create client.conf, etc.). So you can get a similar effect easily, and it doesn't seem that pressing to me to change the seeding now for lucid.
Fine for me for maverick.
| Daniel Richard G. (skunk) wrote : | # |
Remove startup symlinks? I think you mean "update-rc.d cups disable"! And I've done this, to at least not have cupsd running. Just having cupsd installed, however (along with all the printer drivers and supporting infrastructure) is a problem for large sites.
You can remove cups itself, but this then removes the -desktop metapackage (so it's not around to help dist-upgrades, and there's the autoremove problem too) and you still have to get it on in the first place in order to make use of the metapackage at all.
Can we at least get this into lucid-proposed? The issues I found were no worse than you'd get with an unconfigured MTA.
| Martin Pitt (pitti) wrote : | # |
Daniel Richard G. [2010-04-08 17:28 -0000]:
> Remove startup symlinks? I think you mean "update-rc.d cups disable"!
Well, same thing :-) It's usually more obvious to just remove
rc2.d/S50cups, but if you knwo update-rc.d disable, that's fine of
course.
> You can remove cups itself, but this then removes the -desktop
> metapackage (so it's not around to help dist-upgrades, and there's
> the autoremove problem too) and you still have to get it on in the
> first place in order to make use of the metapackage at all.
Right, that's exactly my concern. We get many bug reports where people
have weird bugs, and have ubuntu-desktop uninstalled, and reinstalling
it magically fixes everything. If cups stops being a dependency, this
"get back to a supported configuration" doesn't work reliably any
more.
> Can we at least get this into lucid-proposed?
No, we won't change the structure post-release. If this is done at
all, then pre-release, but I'm still not fully convinced that it
wouldn't create more potential problems than it solves.
Martin
--
Martin Pitt | http://
Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org)
| Daniel Richard G. (skunk) wrote : | # |
Martin, no one can prove a negative. I did concrete testing, and found very little that would be a issue in practice. If a user reports a problem, we can just tell them to create /etc/cups/
What more do you need? Yes, we know that s-c-p doesn't do client.conf right now. How many users do you suppose are going to get their knickers in a bunch over that, after they explicitly overrode the metapackage Recommends: to get there in the first place? How do you weigh that against the complexity and confusion of having both a local and remote CUPS server on a machine that is part of a large LTS deployment?
If there's something more you want me to test, I can do that. Some specific scenario you're worried about, I can try it. I'd like to convince you---because there are clear practical benefits to this change, and it would certainly make *my* job easier---but you're not giving me a whole lot to go on here.
| Martin Pitt (pitti) wrote : | # |
I already said that I'm fine with the change in general. Just that I would not want to change the seed structure of lucid that close to release (this applies to any kind of seed change, not just this one in particular) unless there is a really pressing reason to do so.
| James Westby (james-w) wrote : | # |
Hi,
I'm going to reject this as it is targeting lucid. If the change is still wanted
it could be proposed for precise.
Thanks,
James
Unmerged revisions
- 1468. By MarcRandolph on 2010-03-31
-
cups/foomatic/
openprinting: depends -> recommends LP: #302272

Uncharted territory for me, so this is untested.