[ Home | TOC | Search | Post | Reply | Next | Previous | Up ]
From: Ray Blackwell
Remote Name: 188.8.131.52
Date: 23 Jun 2000
Just a couple of points, it would seem that if the filing system is so 'fragile' that it is impossible for a READO or XLATE to access the data, then the reliability factor is just about zero. To me the real problem lies in the fact that these errors can occur at all, I personally fail to see how the 'Write' computer can create a GFE because of a corresponding READO process. I could understand, only just, if a soft GFE occurred on the 'Read' machine, but the other way round makes the mind boggle.
I have had the same GFE situation occur on 'Indexing' files of a single user system during normal system updating, how on earth does that occur. I realise that 'locking' does not occur on single user machines, but the indexing function is part of the operating system and one would have thought that they could get that right, even if they don't get much else.
On another note, I read your comments on the Sizelock feature of the Linear Hash files and have a question. I have been in the habit of setting large file parameters and then setting sizelock to 2, thereby preventing the file shrinkage from occurring. This setting is also supposed to prevent file expansion I believe, haven't noticed it yet, but I was interested in your sizelock setting of 1, which was I believe intended to prevent shrinkage but allow expansion. After testing this on several files what appears to happen is that the sizelock is constantly reset to zero, thus allowing the file to shrink on the next select, clear etc. This appears to be a little at odds with what was stated about sizelock 1, can you elaborate for me.