--On Tuesday, October 21, 2008 09:21:36 -0400 Dan Williams
<[EMAIL PROTECTED]> wrote:
¦ > Inelegant though it is, i think NM does need a way to define custom
¦ > init (& de-init?) strings externally. This has been the bane of
¦
¦ God no. It's like adding text boxes for arbitrary command line
¦ arguments for openvpn. How are you supposed to know what goes in there?
¦ Do you just google around and find something that some random user
¦ reported worked for them and then try it? Or if that doesn't work try
¦ something else? Then you've just wasted 4 hours for something that
¦ should "Just Work". We don't need to do this, because most of the
¦ differences between hardware can be handled with quirks. Adding an
¦ arbitrary "enter some command strings" means you've completely lost.
¦ Mobile broadband parts are a lot more standardized than
¦ "Hayes-compatible" modems.
i'm not saying it should be an up-front configuration for the user, but
neither should it be embedded in C source. I agree that the general
protocol has been well-established, but there are clearly issues with
initialisation, as this thread has shown (for 2 cases). Even if these cases
are dealt with in source, what happens when a future modem needs something
else a bit different? Do you change the init string in source, then
regression test against every modem that worked before to make sure
nothing's broken, then wait for it to be pushed out to the different
distros? It seems to me an external resource file, pre-built for all known
cases, but editable by "expert" users and/or quickly updateable
independently of the whole package, is much more maintainable.
It's how I'd do it anyway ...
Cheers, Rick
_______________________________________________
NetworkManager-list mailing list
NetworkManager-list@gnome.org
http://mail.gnome.org/mailman/listinfo/networkmanager-list