Mailing List archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[vdr] Error recognition system
Am Freitag, 5. Juli 2002 20:09 schrieben Sie:
> Christian Berger wrote:
> ...
>
> > Ohh FAQs won't help, what we need is an expert system finding out the
> > cause of all errors and giving meaningfull advice.
> > Wake up, the 60s are over! Nowadays computers are powerfull enought to
> > maintain themselves.
>
> I do not believe that what you propose is realistic.
> However, I am sure you will write such an expert system
> and prove me wrong. :-)
OK sure :)
But how about some "context sensitive help" or hints or something.
It shouldn't be hard to run a little procedure at "unusual" situations.
Imagine the card cannot sync to a station. A little "script" could be called
which looks throught all the recent log-entries and then could make certain
statistics, like signal strength by satellite, signal strength by
polarisation and so on. Then it could tell the user where the problem is. For
example if horizontal channels don't sync recently it could tell the user
that there is a problem with the cable.
If we had a little script parser which could get access to some data about
"previous events". We could write simple scripts for every problem we've
found on the mailinglist.
I would recommend scripts because of the following reasons:
Independend of binary file (can be exchanged independendly)
More powerfull commands
save execution (no access to files, etc) makes it save for online updates
How about a syntax like this:
<event> <condition>
Example:
cable_defective nosync(H) nosync(V) >
You'll notice I'm using IPN for the condition. I believe that's easier to
parse and execute.
The Programm would now load all lines in the file and mark them all "unread".
Now it would start at the first line and tries to evaluate it. If another
line is found inside the condition, this line get's evaluated. Every
evaluated line get's marked "read". The results of those lines will also be
stored in case they are needed again. The results can not only be true or
false (1 or 0) but also something in between. For those things we can use the
normal fuzzy logic rules.
In the end we can present the user with a list of the most probable things.
What do you think?
Servus
Casandro
Home |
Main Index |
Thread Index