On May 8, 2009, at 5:19 AM, Michael Schroeder wrote:

> On Fri, May 08, 2009 at 10:21:48AM +0200, Ralf S. Engelschall wrote:
>> This way the whole version comparing stuff is fully pluggable, as one
>> can define how a version is parsed into its chunks and second how  
>> those
>> chunks are compared.
>
> Good luck with providing rpms for multiple distributions that use
> different version comparision settings...
>

You've missed the usage cases, and are assuming that there is
some remaining benefit in the fiction that versioning
applies to _ALL_ packages.

Show me any linux distribution that is attempting mix-n-match
binary packaging across distribution domains. There aren't any.

So version comparison is already split into fragmented
vendor-specific application domains.

And there is no further reason (imho) to pretend that a "universal"
version comparison is important or relevant.

Nor does making rpmvercmp pluggable preclude using a "universal"
scheme if such a scheme is ever deemed necessary.

> Having a fixed algorithm in place also has some benefits.
>

To whom? I have seen no interest from anyone in establishing
a "fixed", as in "standard", not "de facto", version comparison
after discussing the issue of rpm EVR comparisons for *years*.

See my comments to LSB in December for my most recent (and likely last
ever) effort to achieve a "fixed" package version comparison.

73 de Jeff