RFR: more FC4 Requests

[Jeff Johnson]

Panu Matilainen wrote:

>
>For Extras it's irrelevant but I think JBJ was talking about adding apt-
>rpm to Core.
>

I was talking about adding to FC4, and there still appears to be 
interest, although perhaps in passing,
considering the irony that I am suggesting adding apt-rpm to FC4 and you 
are suggesting unnecessary ;-)

Implementing multilib in a depsolver is not that hard, one needs to 
check Provides: and Requires:
color, and match iff the same color for each (or no color, that is the 
traditional rpm behavior). There were
perhaps 10 places that required about 5 lines of code in rpmlib to make 
that happen. The harder problem
was attaching dependencies to files that carry a elf32/elf64 color, but 
that is not an apt-rpm issue.

There are a number of cheats to use repomd data only during initial 
package discovery, just like yum itself
is doing. Alas, a Header is still the only ticket that 
rpmtsAddInstallElement() will punch for a ride through
rpmlib, so yum is pulling the header using a HTTP GET with byte-range 
from the package, and the xml
metadata is used to determine which headers (and packages) need to be 
pulled. Similar things could
be done for apt-rpm.

But I'm not at all sure how and when lack of multilib and repo-md 
support became critical deficiencies
for depsolvers. There are many who love apt still, Fink on Mac OSX comes 
to mind.

73 de Jeff