Re: Reliability of RPC services

[Pierre THIERRY]

Scribit Jonathan S. Shapiro dies 25/04/2006 hora 12:25:
> The watchdog is appropriate where (a) no user is available or (b) we
> may want to point out to the user that a problem exists.

I came to think that we should also consider an opportunity to use
watchdogs without some of their classical drawbacks: someone pointed out
that we can't really expect the user to know what to do when something
goes wrong.

Two notes:

 - When I open a file and I understand after some time that some device
   is not responding, I just kill the application in my shell or WM and
   I'm happy. The average user won't. Maybe he will panic, at worse.

 - When I do anthing on a system under heavy load, know that eventually
   it will complete, remain patient and see the operation time out, I'm
   bothered.

If somewhere a watchdog would notify the user that an operation *seems*
to have timed out, and give an opportunity or instruct how to cancel the
pending operation, everyone could be happy.

If the user wants to wait, he can. If he doesn't want to bother, he also
can, and is given the power to, without prior knowledge.

That is at least the way KDE deals with processes not responding. When
you ask to close a window via KDE (in the title bar, not in the
application itself) and the process does not answer to the signal, a
dialog pops up, explains what is going up and gives the opportunity to
kill the entire application.

Those could be named informative or optional whatchdogs. I don't know if
it is already a well-known design pattern, but it seems it has not been
considered in the current discussion.

Alternatively,
Nowhere man
-- 
address@hidden
OpenPGP 0xD9D50D8A

signature.asc
Description: Digital signature