[Rpm-metadata] files listings

[seth vidal]

On Wed, 2003-10-22 at 11:58, Joe Shaw wrote:
> On Wed, 2003-10-22 at 00:39, seth vidal wrote:
> > but how to uniquely id the packages - should we use the package
> > checksum? or would nevra be sufficient, or maybe just the filename as it
> > is listed in the metadata file.
> 
> I think nevra is sufficient.
> 
> > <package someidentifier attribute>
> >   <file>/some/file/name</file>
> >   <file type='dir'>/some/dir</file>
> > </package>
> 
> I personally prefer this one... feels cleaner and there isn't a syntatic
> difference between directories and files.

I could agree with that - I was more or less listing a couple of options
that came to mind - one thing Jeff L. mentioned was adding more
descriptors for devices and what not - that might get heavier but I
could see some merit in knowing more info about the file - I think it'd
be important to see how heavy it gets before agreeing to it but still
might be worth considering.

> On a related note (and one that we've run into in Red Carpet): How do we
> deal with files that are in an unspecified encoding?  In RC we just drop
> any files which aren't UTF-8 encoded, betting on the fact that probably
> no RPMs depend on one of those files.  Hasn't bit us yet, and seems like
> it'll be less of an issue going forward as distributions move all this
> to UTF-8.  (On a related note, it'd be nice of RPM 4.3 or whatever
> enforced UTF-8 filenames, as it is The One True Way)


hmm - I hadn't considered this.

Daniel: could you take a gander in your rpmfind db of rpms and see if
you can find any non-utf8 filenames? I figure you have the largest
collection of rpms, especially those from non-en_US countries.

-sv