On Feb 23, 2011, at 2:05 PM, Tobias Gerschner wrote:
> Ah could've saved me quite some writing there. The question remains. Any interest to review the current repo-md format and improve it or start over?
These days I'm pursuing pogoplugs running rsync
to get LAN repositories on gigabit ethernet.
If interested, I have 2 (of 3 soon to become 5) pogoplugs
wired up at
for ssh access on NAT'ed ports. Send mail privately if you
want to hack a bit (plugbox linux <- from archlinux currently.
I'd like to see an RPM based ARM distro somehow instead).
The whole idea of mirrored *.rpm and repodata/* is last century now
that NoSQL! databases can be set up in a weekend. The major
problem is the format of the markup is really really hard
to change in intelligent ways. Any change to the markup creates
The other flaw with most depsolvers is trying to do *everything*
while some developer with a short attention span and an itchy ^C
finger is watching the progress bars. Wotta waste of talent imho.
Instead the store-and-forward piece of depsolver's needs to
be split from the depsolving piece. rsync is far more efficient
than the complexity of deltarpm's etc ...
But the simple answer to your question is in
on boxes running @rpm5.org code.
The contents of that file is know as a "queryformat template"
and that's all it is, a --queryformat (from Hell ;-) that can be
adjusted to taste to create whatever repository metadata spewage
one wishes to invent.
You can see several (perhaphs Frankensteinian abominations ;-) work-in-progress
creations of my own in /usr/lib/rpm/qf/*. I need server side infrastructure
like mongo.rpm5.org ro exist before I get to better markup spewage.
73 de Jeff
Received on Wed Feb 23 20:44:21 2011
- application/pkcs7-signature attachment: smime.p7s