Re: dpkg modification: non-interactivity
Wichert Akkerman <wakkerma@cs.leidenuniv.nl> writes:
> Content-Type: text/plain; charset=us-ascii
> Content-Transfer-Encoding: quoted-printable
>
> Previously Brederlow wrote:
> > But what should be done is that when a script falls into interactive
> > mode, e.g. has to ask a question, and thus waits for the user, it is
> > suspended and put into a queue of scripts waiting for user input.
>
> There should really be only one moment during which the installation
> process is allowed to be interactive. Making other parts interactive
> does not make sense if you do things right.
>
> Wichert.
Yep, interactivity should be a short stage, but some programm will ask
a question and then go munching data for several minutes. Then the
next packages comes up and asks a question and goes munching data for
2 hours (well, presume a real slow computer). My proposal is that all
questions are put into a queue, so that when one question is answered
the next one will be there already. Also it would be possible to lay
of questions until later or until somebody is back how knows. That
would allow to leave say cvs unconfigured until you got irc running
and can ask.
Also the aproach with queues wouldn't need a big change in the
scripts. Packages that asks questions would be suspended while
packages that don't can go on. At the end only packages that ask
questions will be unconfigured and then the user can answere to them
when he comes back from cinema and be done with the installation
quickly.
I know that it would still take some time to run the rest of the
script after answering the questions and also further questions will
pop up, but it would be a simple and easy way to improve things by
much. Also, when kept flexible, it then can be changed easily to a
non-interactive aproach via database or other proposals. They can work
fine together.
May the Source be with you.
Goswin
Reply to: