Brock Pytlik wrote:
Jonathan Edwards wrote:
On Aug 25, 2009, at 10:00 PM, Brock Pytlik wrote:
no silly .. i mean the assumption that "dependencies are kept
'correctly' in package metadata" .. also assuming the notion that
"correct dependencies" exist
Isn't this more or less arguing against the idea of packaging
software at all?
nope .. it's a form of protection against the idea that package
creators will *always* "do the correct thing"(tm) and that defined
package dependencies are *always* absolutely correct regardless of
what an administrator might know about the system he's trying to
install .. nobody's saying that we should do away with dependencies
.. just define a way to over-ride them if necessary.
As we've said, repeatedly, there are ways to deal with broken packaging.
But for most sysadmins they are too much time consuming and cymbersome.
Most sysadmins these days have never packaged anything if they hit a
production issue where they will need to install some package without
its dependencies they won't set-up their repository, play with spec
files, get a package rebuilded,.... it just won't happen. Either they
will be able to do it with pkg or they will find another workaround
while cursing Solaris again that it is not as easy as Linux...
And whatever they come up with will almost certainly be even worse than
--nodeps, definitely less managable. --nodeps would provide some audit
entry (history) that it did happen, still a package could be pkg fixed
later on, etc.
What I, and I think others here, continue to be baffled by is why
setting up a repository is such a painful thing to do. Why would you
roll out a whole distribution, no one's ever proposed that. We've
said, "republish the packages you don't like." Could we make the
process simpler? Sure. Shawn's suggested several ways we might be
willing to do that.
Even if there were --force and --nodeps admin still could do it via
modifying a package and publishing it to local repository - that would
always be an option.
But most sysadmins would not even consider it, due to lack of skills,
too much overhead, etc.
Especially when you need to fix something right now and you've never
played with pkg repositories, spec files, etc.
Btw, if all you want to do is put the files down, then back them out,
zfs snapshots are wonderful things. It only provides one axis of
change, but with some scripting love, I'm sure you could do exactly
what you want without ever having to deal with the packaging system at
all. I still don't get how that's different than -zforce, except it
doesn't live in the packaging tool (because it breaks the notion of a
well formed system).
I don't like such attitude.
yes, and if you don't like pir patching tools write your own, and we
(well, mr, Brown) did - see pca.
then if you don't like lack of automation in our packaging tools
(downloading and installation with dependencies) then come up with your
script - and we (Blastwave) did - pkg-get.
With such an attitude why do you bother with developing pkg at all?
I'm sorry, but really - the whole discussion has little to do with
technicalities as most of us agree it is possible to implement, the
discussion seems to be more about persuading people who have been doing
--nodeps like workarounds for years in production that they actually
shouldn't because all packages we (as Sun and community) going to
provide will be perfect and you will never need such a feature, and if
you need to then write your own tools. I'm sure it will happen at some
point but really such a feature should be a part of packaging tools as
it is on most other platforms, at least on one of the most popular these
days - Linux.
--
Robert Milkowski
http://milek.blogspot.com
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
