On Tue, 2007-10-23 at 23:32 +0100, Adam Huffman wrote:
On Tue, 23 Oct 2007, Dan Williams wrote:
> Right; NM 0.7 and later is definitely _not_ targetted at a "very narrow
> niche". It's targeted at hopefully >= 75% of the use-cases of Fedora.
> Perhaps this weekend (unless I'm still fixing bugs with NM in general)
> I'll integrate the work Soren did for a system config service which will
> allow connections to be brought up before you log in. That's the only
> real blocker to using NM in a GUI-less server-type environment if you
> like. Other than that, most of the bits for static IP support are in
> (applet bits to follow), the options to lock things like 802.3 duplex,
> speed, and other stuff are in-progress by Tambet, and most of the
> wireless options work.
> By the holidays, I expect we can flip NM on by default in Rawhide, even
> if it's not default in F8.
A general comment re. NM. It did seem odd the way such a large change
was made so late in the F8 development cycle, with no real discussion I
could see (please do point me to such a discussion if there was one).
It's mostly about the D-Bus interface. Since the D-Bus interface is no
longer insane, but quite clear and usable, switching to 0.7 as an F8
update is completely out of the question. With the new architecture,
it's possible to add the following things as follow-on updates to F8
that would be impossible using 0.6.x branch:
1) multiple active devices
2) good static IP support
3) PPP and Bluetooth support
4) Broadband card support (GPRS/UMTS/EVDO/etc)
5) internet connection sharing
6) better handling of similar APs with different security or on
different radio bands
These are all things people want and all are only possible building on
the architecture that 0.7 provides.
Yes, it's a gamble, and it'll be a bit rocky. But that's _also_
compounded by landing the mac80211 stack and it's 8 or so new drivers in
F8 as well (p54, iwl3945, iwl4965, b43legacy, b43, rt2x00, ath5k,
zd1211). So many bugs people might encounter with wireless in F8 will
be of kernel driver origin, not necessarily due to NetworkManager. We
just have to sort out the bugs wherever they may be.
For quite a long time it was unusable for lots of people and still
a little flaky.
Was it a case of the rewrite taking longer than planned?