Hi,

(could you please keep 527...@bugs.debian.org cc'ed so that the bugreport 
subscribers can read us ?)

Le vendredi 15 mai 2009 22:05:39, vous avez écrit :
> Didier Raboud schrieb:
> > My idea was not to have a dialog initiated by a hal .rules but to have a
> > simple wrapper which any user can launch. This wrapper could have an
> > option like --non-interactive which could do the actual switch if the
> > DV:DP:UUMID is unique and do nothing (return error) if not.
>
> I'm not sure I understand that fully ...
> How would a user know he is supposed to start the wrapper? And does
> he have to do it every time he plugs his device in?

It would work as now, with two possible cases:

1) automatic launch by a udev rule
2) hand launch for a not-yet-recognised device or without udev rules

If the device doesn't have any rule for it, the user will have to launch it by 
hand (as actually without my additions in the Debian package), but if the 
device 
is known to usb-modeswitch (by an appropriate file in the database), the 
wrapper 
will be launched with --non-interactive.

> I still like the idea of a "run-once-per-device installer" better;
> if I remember correctly, a similar approach is chosen by the SANE
> scanner project.

I dislike it, because it transfers the burden of configuration from 
upstream/distribution (aka you/me) to the user, where we could make him an 
easier life.

But I can live with it, you are upstream afterall. :)

> > What are your plans for future releases ? Doing the split of the
> > configuration files could be a good first step.
> > 
> > Then, the detection-and-prompting wrapper can be written in simple bash
> > or something similar and I can handle the auto-detection .rules file (as
> > now).
>
> I would prefer a tcl script which is my favourite "rapid
> development" language. A tcl interpreter/shell is likely installed
> on most systems and is small enough to quickly add it where it's
> missing. The readability is better than Bash or Perl and it's an
> excellent tool for text parsing and "glueing" together common
> command line tools.

I can live with that too. Your choice !

> The USB_ModeSwitch tool should not do much more than it is doing
> now; I'd rather not add a lot of complexity. If the wrapper /
> installer mechanism proves to demand more options I would certainly
> implement them, but apart from that I see the further development
> outside of the binary.

I agree. The transfer to a simple "database" infrastructure seems to be 
reasonably easy with the existing binary - further improvements seem to have to 
be outside of it.

> Cheers,
> Josh

Regards, 

OdyX

-- 
Didier Raboud, proud Debian user.
CH-1802 Corseaux
did...@raboud.com

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to