[Rpm-metadata] take 3 (or is it 4) whatever :)

[seth vidal]

Hi,
 I read through all the discussion of namespaces and orentries and
things that have come up in the last couple of days.

I think Joe has a point, unless the goal is to converge the formats
(which I don't think anyone is doing atm, and is a whole other kettle of
fish anyway) then it seems like we're only represening either a .deb or
a .rpm or something else.

so let's say I have a dir that is full of debs AND rpms - I run a
command to index them and output them in this format.

now the tool that comes along behind will need to know if they are rpms
or debs or what. So we can't homogenize the format so much that there is
little way to tell which is which.

So maybe a simple thing to start with:

<package xmlns="http://www.yoursite.org/ns/rpm">
<name>foo</name>
<arch>i386</arch>
<arbitraryns:some-metadata />
....
</package>

So I'm taking part of what Darrin recommended but giving the parser a
quick way to skip the package if they don't understand it/don't need it.

you check the namespace for the <package> - if it is not one you can
deal with then you move on.

Then we can get into the nitty-gritty of defining the namespace for each
of the formats.

And the rules get simpler b/c then all the package mgmt tools have to do
is know what things they can and cannot deal with.

so, for example, if jbj gets his way and rpm becomes capable of dealing
with debs, then I can make yum start grokking the :
<package xmlns="http://thatsite/ns/deb">

but until then I skip the ones that are not rpm.

then you can index huge dirs of both deb and rpm and anything else and
keep from going crazy on the pkg mgmt side.

then we're right back to the metadata representation but w/i segmented
spaces.

maybe this is not attractive, I dunno, but it means its a lot faster to
get through a lot of data to get to the stuff you can use.

-sv