Yes, I'm not saying we do it right.  I'm just saying we correct things
as we find them.   I know of at least one instance where I was hit by
something like this and the fix used was to detect the error at the
initial point of failure.

On 8/3/07, Kevin Menard <[EMAIL PROTECTED]> wrote:
> > -----Original Message-----
> > From: Mike Kienenberger [mailto:[EMAIL PROTECTED]
> > Sent: Friday, August 03, 2007 11:59 AM
> > To: [email protected]
> > Subject: Re: Exceptions . . .
> >
> > In my opinion, it's best to trap the error as soon as possible.   As
> > far as I remember, that's our standard practice, and as we come across
> > items like this, we correct them.
>
> I may be mistaken, but my experience tracking down some problems in the
> past couple days indicates that we rarely check that invariants hold
> true.  Instead, things go as far as they can, the exception is caught,
> wrapped, and sent back up the pipeline.  I imagine it's been done this
> way as a performance benefit, but I'm not sure the overhead would be
> that great anyway.  Comparisons are pretty cheap.
>
> --
> Kevin
>

Reply via email to