I propose new beta criterion:
"The installer must be able to install system with minimal usable set of packages."
There is test case [1] associated with alpha, I'd change it to beta release level.
[1] https://fedoraproject.org/wiki/QA:Testcase_Package_Sets_Minimal_Package_Inst...
On Mon, 2012-01-30 at 11:25 -0500, Petr Schindler wrote:
I propose new beta criterion:
"The installer must be able to install system with minimal usable set of packages."
There is test case [1] associated with alpha, I'd change it to beta release level.
[1] https://fedoraproject.org/wiki/QA:Testcase_Package_Sets_Minimal_Package_Inst...
We could phrase it a little bit more similarly to the existing default install criterion:
"The installer must be able to complete package installation with the default package set for each supported installation method"
Perhaps:
"The installer must be able to complete package installation with a minimal usable set of packages"
On Mon, Jan 30, 2012 at 7:38 PM, Adam Williamson awilliam@redhat.com wrote:
We could phrase it a little bit more similarly to the existing default install criterion:
"The installer must be able to complete package installation with the default package set for each supported installation method"
Perhaps:
"The installer must be able to complete package installation with a minimal usable set of packages"
Not to be pedantic, but what's the difference in these two? Are we actually suggesting that the default package set is *not* useful? :). I get what Petr is going for, but either of these wordings don't get it across to me. What I consider a "minimal usable set of packages" and what you do may be two entirely different things. What I would rather do is specify the minimal amount of function that the system must be able to perform, and to me, that reads something like:
"The installer must be able to install the minimum amount of packages required to obtain network connectivity via any supported means and install additional packages as the user sees fit"
Yes, I realize that this criterion, when applied, means that we can't install things like vim or openssh during such an installation, and that's fine with me. Remote access is not a requirement, nor is the ability to edit files (but if we want to add them, that's fine - we just need to explicitly define what tasks are able to be performed or not performed)
On Mon, 2012-01-30 at 19:57 -0500, Jon Stanley wrote:
On Mon, Jan 30, 2012 at 7:38 PM, Adam Williamson awilliam@redhat.com wrote:
We could phrase it a little bit more similarly to the existing default install criterion:
"The installer must be able to complete package installation with the default package set for each supported installation method"
Perhaps:
"The installer must be able to complete package installation with a minimal usable set of packages"
Not to be pedantic, but what's the difference in these two? Are we actually suggesting that the default package set is *not* useful? :).
The first one is for the 'default' package set - the several GB you get from the DVD, including X and GNOME and so on. The second is for the 'minimal' package set.
I get what Petr is going for, but either of these wordings don't get it across to me. What I consider a "minimal usable set of packages" and what you do may be two entirely different things. What I would rather do is specify the minimal amount of function that the system must be able to perform, and to me, that reads something like:
"The installer must be able to install the minimum amount of packages required to obtain network connectivity via any supported means and install additional packages as the user sees fit"
Yes, I realize that this criterion, when applied, means that we can't install things like vim or openssh during such an installation, and that's fine with me. Remote access is not a requirement, nor is the ability to edit files (but if we want to add them, that's fine - we just need to explicitly define what tasks are able to be performed or not performed)
Yeah, this is kind of problematic, because I don't really want the release criteria to prescribe exactly what the 'minimal' package set should include. Perhaps we should just explicitly refer to 'the installer's "minimal" package set' or something like that.
On Mon, Jan 30, 2012 at 8:49 PM, Adam Williamson awilliam@redhat.com wrote:
Yeah, this is kind of problematic, because I don't really want the release criteria to prescribe exactly what the 'minimal' package set should include. Perhaps we should just explicitly refer to 'the installer's "minimal" package set' or something like that
True - come to think of it, I really don't believe that it is QA's domain to define the task list - but it should be somebody's. What I don't want is the feature creep of the "minimal" package set - next thing you know, GNOME will be part of that package set. But I don't think that QA is the appropriate place to address those concerns :)
On Mon, 2012-01-30 at 21:56 -0500, Jon Stanley wrote:
On Mon, Jan 30, 2012 at 8:49 PM, Adam Williamson awilliam@redhat.com wrote:
Yeah, this is kind of problematic, because I don't really want the release criteria to prescribe exactly what the 'minimal' package set should include. Perhaps we should just explicitly refer to 'the installer's "minimal" package set' or something like that
True - come to think of it, I really don't believe that it is QA's domain to define the task list - but it should be somebody's. What I don't want is the feature creep of the "minimal" package set - next thing you know, GNOME will be part of that package set. But I don't think that QA is the appropriate place to address those concerns :)
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
From: "Adam Williamson" awilliam@redhat.com To: "For testing and quality assurance of Fedora releases" test@lists.fedoraproject.org Sent: Tuesday, January 31, 2012 4:51:56 AM Subject: Re: New criterion for installation with minimal set of packages
On Mon, 2012-01-30 at 21:56 -0500, Jon Stanley wrote:
On Mon, Jan 30, 2012 at 8:49 PM, Adam Williamson awilliam@redhat.com wrote:
Yeah, this is kind of problematic, because I don't really want the release criteria to prescribe exactly what the 'minimal' package set should include. Perhaps we should just explicitly refer to 'the installer's "minimal" package set' or something like that
True - come to think of it, I really don't believe that it is QA's domain to define the task list - but it should be somebody's. What I don't want is the feature creep of the "minimal" package set - next thing you know, GNOME will be part of that package set. But I don't think that QA is the appropriate place to address those concerns :)
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
So new beta criteria should be:
"The installer must be able to complete package installation with a minimal usable set of packages"
And what "minimal set" means should be defined somewhere else? And as soon as it will be somewhere, we should give the link.
Petr Schindler (pschindl@redhat.com) said:
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
So new beta criteria should be:
"The installer must be able to complete package installation with a minimal usable set of packages"
And what "minimal set" means should be defined somewhere else? And as soon as it will be somewhere, we should give the link.
@core
Bill
On Tue, 2012-01-31 at 10:52 -0500, Bill Nottingham wrote:
Petr Schindler (pschindl@redhat.com) said:
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
So new beta criteria should be:
"The installer must be able to complete package installation with a minimal usable set of packages"
And what "minimal set" means should be defined somewhere else? And
as soon as it will be somewhere, we should give the link.
@core
That's an implementation detail. It's not a capability-driven description of which packages should actually be in the minimal package set, as was discussed earlier in the thread.
Adam Williamson (awilliam@redhat.com) said:
On Tue, 2012-01-31 at 10:52 -0500, Bill Nottingham wrote:
Petr Schindler (pschindl@redhat.com) said:
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
So new beta criteria should be:
"The installer must be able to complete package installation with a minimal usable set of packages"
And what "minimal set" means should be defined somewhere else? And
as soon as it will be somewhere, we should give the link.
@core
That's an implementation detail. It's not a capability-driven description of which packages should actually be in the minimal package set, as was discussed earlier in the thread.
Merely stating that if you're linking to what the minimal set of packages will be, that's it.
Bill
On Tue, Jan 31, 2012 at 2:56 PM, Bill Nottingham notting@redhat.com wrote:
That's an implementation detail. It's not a capability-driven description of which packages should actually be in the minimal package set, as was discussed earlier in the thread.
Merely stating that if you're linking to what the minimal set of packages will be, that's it.
But to Adam's point, who defines what is in @core and what it can do? Could I decide tomorrow that the GNOME desktop is a core functionality of the distro and commit it to comps and so it is (I seriously hope someone would come shoot me if I *actually* did that :) )?
I guess this goes to the point that no one "owns" comps groups, but I think someone should, and @core (and to a lesser extent @base) should be "special" to have some defined set of functionality.
But I think this is getting *way* off topic for QA and should be a fesco discussion :)
On Tue, 2012-01-31 at 10:52 -0500, Petr Schindler wrote:
From: "Adam Williamson" awilliam@redhat.com To: "For testing and quality assurance of Fedora releases" test@lists.fedoraproject.org Sent: Tuesday, January 31, 2012 4:51:56 AM Subject: Re: New criterion for installation with minimal set of packages
On Mon, 2012-01-30 at 21:56 -0500, Jon Stanley wrote:
On Mon, Jan 30, 2012 at 8:49 PM, Adam Williamson awilliam@redhat.com wrote:
Yeah, this is kind of problematic, because I don't really want the release criteria to prescribe exactly what the 'minimal' package set should include. Perhaps we should just explicitly refer to 'the installer's "minimal" package set' or something like that
True - come to think of it, I really don't believe that it is QA's domain to define the task list - but it should be somebody's. What I don't want is the feature creep of the "minimal" package set - next thing you know, GNOME will be part of that package set. But I don't think that QA is the appropriate place to address those concerns :)
Yeah. As far as QA is concerned, the key questions are 'is there a minimal package set present, does an install with that package set complete properly, does it boot'. What's *in* it is not really our concern.
So new beta criteria should be:
"The installer must be able to complete package installation with a minimal usable set of packages"
And what "minimal set" means should be defined somewhere else? And as soon as it will be somewhere, we should give the link.
More or less, yeah. Maybe we should make the criterion:
"The installer must offer a 'minimal' installation option, and must be able to complete installation successfully when this option is selected"
(note that a later criterion specifies that any install according to earlier criteria should boot, so we don't need to specify that).