If you're going to throw more fuel on an already terse disagreement, please bother to spend the time actually reading the arguments in question.
This has absolutely nothing to do with the merits of having apt open STDIN. The fact of the matter is, that's the way apt/dpkg works today and no one has bothered to start the process of changing that policy. Yes, it's a bad practice. No one disagrees with that. But if no one working on PackageKit has bothered to even ask that the practice be changed, then the apt2 backend is going to be forever stuck at the stage it's at now: an interesting proof-of-concept.
It takes time and cooperation to get to a common practice that PackageKit can support. If no effort whatsoever is made to cooperate--ie. a unilateral declaration that apt's design will not be tolerated--then yes, that's myopia.
Update 2008-04-14 11:46 PM: A poster who posted nine replies to this and the last post has been banned after being warned not to continue to spam my blog. All comments that display as screened
were by the banned party.

Comments
I think trying to get the buy-in of distributions is the tail wagging the dog; they're not the people who actually matter, the users are. PackageKit is a very nice solution to real problems distros face: e.g., Firefox/Tbird plugins, codec installation, Monodevelop plugins, Gimp plugins. There are all sorts of reasons you want to interface with 'whatever' to install new (parts of) software. It's free software; if a distribution feels they need to change it to make it "work better" in their distro, that's really up to them as much as the original author.
If debian is unwilling or unable to do that then Debian will die. Survival of the fittest.
That being said, I think the entire issue has been blown out of proportion. Does it suck that sometimes Debian-based tools requires input? Yes, of course it does and steps should be taken to elliminate the practice.
Does it happen often? Not really. At least not on desktop-machines. On my Ubuntu-machine I can think of only four
packages that has required it:
1. Sun-java required me to accept the license
2. hddtemp asked me if I wanted it to run as a deamon and on what port.
3. hotsmtpd did the same thing as 2.
4. Google-earth asked me to accept the license (and no, it's not available in the standard Ubuntu repos for obvious reasons)
There may be other packages that requires input as well, but:
1. As far as I can tell, there is work being done to make example 1 and 4 possible using packagekit.
2. Things can be packaged without requiring input. Those (few in my experience) that now do require it could be repackaged, and I would assume that can be done without explicitly requiring a change in Debian policy.
See a pattern?
Fernando
Or, maybe people don't have enough time to actually do anything more than skim the articles that they think that disagree with. Nah, that's just silly...
Call it human nature, call it whatever, but even if you are correct, because of the manner of your argument, you have made yourself no friends here.
This is because the graphical app you see on your desktop is not the one running apt or dpkg: that happens in a backend process running as another user, not even connected to your X server.
Ignoring PackageKit for a moment, don't you think it is a problem that these packages can't be installed or upgraded in an unattended fashion?
Looking at the code, I see that it skips packages that would prompt for conffile changes, but I don't see anything that'd handle arbitrary usage of STDIN (other than the ability to blacklist problem packages).
It seems to be trying as hard as it can to avoid interactions you say are essential.
It's spelled Red Hat, not RedHat or redhat as previously pointed out on a recent blog entry of yours. Amusingly you did correct your blog entry for the error in using the word "infinitesimal" but not for this.
So all this is a little childish if you ask me. But, hey, I work for Red Hat that maybe that makes me the devils goon. Oh, and I should point out that nothing in this message in no way reflects the opinion of my employer.
That said, I'm hoping someone will help sort out this situation so PackageKit and Debian are compatible.
Thanks,
David
There seems to be two statements people are making here:
- PackageKit should support Debian based distributions.
- PackageKit should support the entire range of behaviour that dpkg currently allows.
I think most people want the first, but see the second as a lower priority or feel that it isn't necessary to achieve (1).PackageKit isn't the first thing to have trouble with these problem packages, but it might provide a reason to fix them.
To bring this back to the start of the controversy, it does matter if you write software that embeds packageKit for Hardy (with 2 weeks to go, I don't see that happening, so maybe Intrepid), and it ends up not working on Debian systems. A worse scenario would naturally be where Intrepid does ship packageKit libraries, and the package you wanted tanks your app and leaves dpkg in a half configured state because debconf was set to ncurses and High instead of GTK and Default (or because the package uses STDIN, a bad practice, especially in light of the package-kit design).
- Chris
All the whois I could find was that it's owned by Qwest headquartered in Colorado.
Prompting on stdin (or any other method that does not conform to the Debian Configuration management specification, version 2 or higher) is deprecated under Section 3.9.1 of Debian Policy. Prompting in general should be kept to a minimum. See:
http://www.debian.org/doc/debian-po
PackageKit is thus perfectly justified in saying that packages which block on stdin are broken - they should be updated and fixed. So long as it handles debconf questions, there's no problem.