We have used this method of walking ksam files (CM) for years without
trouble or user reports of problems. We provided first, next,last,back
navigation of Image sets, so doing Ksam was a natural component. The
probable saving grace was that the ksam files were mostly stable with few
additions, and almost non-existent deletions.


jp
----- Original Message -----
From: "Tracy Pierce" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Saturday, April 12, 2003 5:20 AM
Subject: Re: [HP3000-L] FSPACE and KSAM


> > Tracy Pierce wrote:
> >  > I'm not convinced that FSPACE(fn,-1) really does work reliably with
> >  > CM KSAM: How can it?  All the docs I've read (ie not all of them!)
> >  > indicate that the indices are single-threaded, leaving reliable
> >  > backspace theoretically impossible: there is no backward
> > pointer like
> >  > there is on Image paths. Please tell me I'm wrong?
>
> Jeff replied,
> > Originally in CM KSAM there was only one record pointer, period.  If you
> > wanted to manipulate the record pointer, you were supposed to flock()
> > before, do your things, and funlock() afterward.  If you fpoint() or
> > freadbykey(), the record pointer can be changed by a similar operation
> > in another process, and the pointer no longer points to the record that
> > you think you are.  The same would apply to fspace().  You had to lock
> > the file around accesses which were dependent on the record pointer.
> >
> > I think this still applies to KSAMXL but I am not confident of this.
> > Jeff
>
> agreed; you're sharing the index pointer and buffer with other processes,
> and they'll lose your place.
>
> but what about the backspace bit?  If we have only forward-pointing links,
> how is backward possible or even seemingly possible?  Only one answer
comes
> to mind: if you load the file in the same sequence as the key you're
using,
> the physical backspace appears to work.  but what about a different key?
>
> I'd sure love to be wrong about this: it would be just great to really
> backspace, even if locks are required.
>
> Tracy Pierce
>
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *