Re: More package questions
>>>> Olivier Galibert writes:
On Wed, May 06, 1998 at 06:42:06AM -0700, JJ wrote:
> It is only supposed to update the packages you have, and decides
> which to update based on package-get-base. Currently, it may end
> up installing packages you do not have if they are required by
> packages you do have. I think that should be changed since
> sometimes package A is officially required by package B, but there
> are no real problems in running B without A.
In *this* case what we have to fix is the dependency. We should
have a "dependency overrride" flag, but *not* set it by default.
Yes, we should fix dependencies which are not required. But this
should be independent of package-get-update-all. If user A
initially gets packages using package-get-all, he will fetch
dependent packages as well. If user B thinks he is smarter than
the dependency system, he can fetch get only the ones he wants with
package-get (or plain old ftp). Either way, the user is in control of
whether dependent packages are automatically fetched.
Now these two users decide to update their packages with
package-get-update-all. Currently user B ends up fetching the
dependent packages he was intentionally did not download to start
with. I think that conflicts with the principal of least surprise.
If package-get-update-all is (trivially) changed to not go after
dependent packages, then it acts as expected for both user A and user
B. It works for user A because he has dependent packages, and if
package-get-base shows a newer version of one of those has come out,
package-get-update-all will still get that newer version.
John Jones
jj(a)asu.edu