[Rpm-metadata] primary refinement
Jeff Sheltren
sheltren at cs.ucsb.edu
Mon Feb 7 05:22:35 UTC 2005
On 2/6/05 4:39 PM, "seth vidal" <skvidal at phy.duke.edu> wrote:
> So i'm looking at primary.xml.gz and I'm thinking of useful ways of
> refining the data included a bit so it's a bit lighter to read in.
>
> I was thinking what do y'all think of moving the following fields to a
> different xml file in the repository?
>
> summary
> description
> packager
> license
> vendor
> group
> url (url for the package, not url location of the rpm)
> buildhost
>
>
> my thinking is that it would decrease the data in primary.xml.gz and
> make actions that are most common, 'update, install, etc' a fair bit
> faster simply by not having to go through all these nodes.
>
> every package has most of these entries, they're fairly text heavy and
> not useful for solving deps or finding out which packages are available.
>
> thoughts?
>
> -sv
>
This sounds like a good plan to me. The primary.xml file ends up being quite
large when looking at all the packages in a distribution. I did a quick
test by running two versions of createrepo on the FC3 base packages - one
version creating the normal primary.xml file, and the other creating a
trimmed down version without the fields you mentioned. The trimmed down
version ended up being about 25% smaller filesize (by MB, not by line
count). I imagine that should speed things up quite a bit for tasks where
that information is not needed.
-Jeff
More information about the Rpm-metadata
mailing list