Re: Bits from the Release Team - Kicking off Wheezy
Le dimanche 03 avril 2011 à 07:24 +0800, Paul Wise a écrit :
> Probably using glib/gobject etc is a no-no for a package that needs to
> be in base.
And that’s because… ?
> The main problem I see is that the design of NM is wrong™ and the
> upstream maintainers do not see it that way.
>
> The upgrading/restart issue was partially fixed, I read that wired
> connections without authentication are not killed any more.
>
> The design issue is that NM thinks it is the centre of the networking
> universe instead of just part of it, with the kernel at the centre.
This is not a design issue, this is what can make it work. If you have
one single, extensible daemon being able to manage all your connections,
you don’t need anything more. The problem here is that it is, currently,
not being able to handle all the network connection types we want to
handle.
My question remains: is it better to add support for them to existing
software that already does a good job, or to continue masturbating over
vaporware?
> The netconf design was much better here.
Vaporware with good design is still vaporware, unfortunately.
> NM does do a lot of things well though and I've been almost happy
> using it the past few years.
>
> The upgrading/restart issue is just a symptom of that problem, more include:
>
> Changes made to the current network setup manually using
> ip/route/ifconfig are not reflected in the UI.
>
> Hardcoding policy like wired > wireless and killing my default route
> when switching.
These are all valid issues that would of course need fixing - although
for manual ip/route/… I don’t know how this could be implemented in a
correct way.
--
.''`.
: :' : “You would need to ask a lawyer if you don't know
`. `' that a handshake of course makes a valid contract.”
`- -- J???rg Schilling
Reply to: