DNF: why does it refresh metadata all the time

[Dan Williams]

On Fri, 2014-06-20 at 23:27 +0200, poma wrote:
> On 20.06.2014 17:55, Dan Williams wrote:
> > On Fri, 2014-06-20 at 08:55 +0200, drago01 wrote:
> >> On Thu, Jun 19, 2014 at 8:59 PM, Jared K. Smith
> >> <jsmith at fedoraproject.org> wrote:
> >>> On Thu, Jun 19, 2014 at 2:01 PM, Reindl Harald <h.reindl at thelounge.net>
> >>> wrote:
> >>>>
> >>>> if *that* is what is supposed to make DNF faster it's just a lie
> >>>
> >>>
> >>> This is not the only thing that DNF does differently to try to make package
> >>> installations and updates go faster (or appear to go faster).  Calling the
> >>> developers liers doesn't help the situation any.
> >>>
> >>>>
> >>>> if i am really interested in updates now i do "yum clean metadata && yum
> >>>> upgrade"
> >>>> for many years simply because you don't know how accurat you metadata are
> >>>
> >>>
> >>> Sure, but you have to understand -- you're a power user.  You know enough to
> >>> do this in yum for your particular use case, which means you probably know
> >>> enough to change the DNF settings with regards to cron-based metadata
> >>> retrieval.  What I think you're missing (and frankly, seem to miss in the
> >>> lot of fedora-devel discussions you take part in) is that Fedora isn't
> >>> engineered around *your* particular needs.  We do things mostly by
> >>> consensus, and aim to make it a pleasant experience for the *average* user
> >>> (or whatever we have in the Fedora community that approximates an average
> >>> user), and not just for power users with very specific needs and
> >>> requirements.
> >>>
> >>> Whether you like it or not, one of the most common complaints about yum
> >>> (especially from people coming from another package management system) is
> >>> that it seems slow because of the necessity to download the metadata.  The
> >>> DNF developers -- in trying to address this common complaint -- had solved
> >>> it by handling metadata in a different way.  They've also added settings so
> >>> that power users like you and I can tune it to better fit our particular
> >>> needs.
> >>>
> >>>>
> >>>>
> >>>> and *no* traffic is not cheap everywhere, by far not
> >>>
> >>>
> >>> I probably understand this better than a lot of people on this list, as I've
> >>> been on a bandwidth-limited connection for the past nine years.  Only in the
> >>> past month have I been able to get high speed internet in my home that
> >>> wasn't limited to a few gigabytes per month.  So yes, I completely
> >>> understand that traffic isn't cheap (or fast) everywhere.
> >>
> >> It should be at least smart enough to not do it on mobile broadband
> >> (like packagekit does).
> >
> > Python + D-Bus example for detecting WWAN NetworkManager 0.9+ is here:
> >
> > http://cgit.freedesktop.org/NetworkManager/NetworkManager/tree/examples/python/dbus/is-wwan-default.py
> >
> > Dan
> >
> >
> 
> This is super duper, however if wwan is on the router as Ranhald wrote, you can only click your heels three times and repeat, "There's no place like home."

Certainly.  But that doesn't mean we shouldn't try to fix 50%, even if
we can't achieve the stars.  So I think there's a ton of value in doing
this despite the fact that we can't be perfect.

Dan