Hi Michael,

Michael Hunter wrote:
> On Tue, 26 Jun 2007 17:18:15 +0100
> Darren Kenny <Darren.Kenny at Sun.COM> wrote:
> 
> [...]
>> I understand what you're saying, but I think any user would find it strange 
>> that
>> when they load the UI there are value fields being flagged as invalid -
>> generally you would only want to see errors that *you*, as the user, has 
>> made,
>> not what anyone else did, otherwise it'd be quite confusing. Still, there 
>> may be
>> some value to what you suggest, just it could be difficult to handle in a 
>> "user
>> friendly" way, I'll have to think about it a bit more...
> 
> I just dealt with dtpower(1M) being unwilling to load a config or save
> over a broken config.  This is really very unusable.  This kind of
> error handling is tough.  Initially I think being willing to say "dump
> to file and delete or don't load" is fine.  As time goes on though we
> are going to have to develop strategies for dealing with the uglier
> types of broken config we run across.  This is no different then what
> we should do if it was a flat file.

I'm not saying that the UI shouldn't write over a broken configuration - it
definitely should - what I'm saying is that on loading, a "valid" configuration
should be passed to the UI - if the entry in the repository is broken then so be
it - from the UI perspective it would still be valid to overwrite the broken
configuration, since it's quite likely that the user would change things since
it's not currently working.


> 
>> As a question here, how is it proposed that the NWAM daemon would handle 
>> invalid
>> configuration - simply ignore them/report an error and/or fail to start?
> [...]
> 
> Mostly the second.  Possible some of the first.  The third is just so
> unusable as to not even be under consideration.  It is dangerous to do
> the first if you think there is any possibility it creates a security
> problem.

I'm fairly sure that we should at least inform the user that the current
configuration is broken and what they are seeing is a modified/sanitised version
of this configuration.

Thanks,

Darren.


Reply via email to