[packagekit] The dbus backend

[Robin Norwood]

On Tue, 12 Feb 2008 07:47:19 +0000
Richard Hughes <hughsient at gmail.com> wrote:

> On Mon, 2008-02-11 at 22:21 -0500, Robin Norwood wrote:
> > On Mon, 11 Feb 2008 23:28:06 +0000
> > Richard Hughes <hughsient at gmail.com> wrote:
> > 
> > > rnorwood and I have been working on the yum2 dbus backend. I've
> > > actually got the yum2 backend to run now, with search times down
> > > from 3 seconds to 0.7 seconds. So quite a dramatic speedup. Good
> > > stuff to look forward to.
> > 
> > Wait, it works?  Hey, it works!  Neat!  It wasn't working at all
> > for me until I read your message and tried again.  I guess dbus
> > just needed to think about things for a bit.
> 
> I merged a few fixes. I don't pretend to "do" python, but I think I'm
> getting better :-)
> 
> > Ok, I don't get it.  Why keep the text at all if you want ints?  Why
> > not pass everything up the stack as ints, and let the UI translate
> > from int -> string?
> > 
> > enum.py:
> > STATUS_CANCEL = 4
> > STATUS_CLEANUP = 7
> 
> Then we break the spawned backends like yum and conary :-) Hence why
> we need new constants. I'm currently erring at sending text and
> decoding, so people could rely on just a little bit of ABI stability
> without grepping pk-enum.c.

Yeah, but we can generate the enums for all of them fairly easily and
just use the numeric constants.  OTOH, converting in the daemon
shouldn't be any noticeable speed difference either.  No big difference
between the options, except I don't like the STATUS_CANCEL and
STATUS_CANCEL_ID one.  Reminds me of java code. :-)

-RN

-- 
Robin Norwood
Red Hat, Inc.

"The Sage does nothing, yet nothing remains undone."
-Lao Tzu, Te Tao Ching