High water marking

[Wim Van den Wyngaert]

John,

This seems to be the same discussion as in databases. First, only table locking was implemented. Next (on certain systems like HP3000 Image) there was chain locking. And finaly there was row locking.

It seems to me that VMS must block all IO's for the part being erased, not the whole directory or disk.

Sorry, but nobody can convince me to block all IO's.

Hein : it must have been a long queue because 11 seconds is a long time.

Wim

[John Eerenberg · Accepted solution]

Wim,

I've been looking into this. The more I find, the more questions I have. I have found an inconsistent blocking of IO when doing a file create vs. /erase; sometimes the directory, sometimes the whole disk. This should not happen so I have logged a call with HP and the support folks are scratching their heads. ;-)
When I have more, I'll post it.

This impacts our usage of highwater marking and delete/erase and the lexical functions we use as well. So I am glad you raised this question!

john

[Wim Van den Wyngaert]

Thanks John