From: kuhrt@eisner.decus.org Sent: Monday, July 19, 1999 7:36 PM To: Info-VAX@Mvb.Saic.Com Subject: Re: What symbols and logicals are "really" used? In article <009DB57E.E4FA04BF@SendSpamHere.ORG>, system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-) writes: > In article <7mvhfc$425$1@sniff.shr.dec.com>, lederman@star.enet.dec.DISABLE-JUNK-EMAIL.com (Bart Z. Lederman) writes: >> >>The question originally asked brings up an interesting >>general case question, which is: >> >>Should there be a way to enable accounting or usage >>or security audits on logical name translations? >> >>In theory, it ought to be possible to add such a thing >>to the operating system. However, it would almost certainly >>increase overhead on logical name translation when in use, >>and possibly even a slight bit of overhead even when it's >>not in use (the system would at least have to check to see >>if auditing or whatever was turned on). So this would >>lead to additional questions, such as: would a large >>enough portion of the user base actually use this feature, >>so that it would be worth while allocating engineering >>resources into developing it (which means that they would >>not be available to work on something else)? Would it be >>worth the extra overhead? > > Likely not. > >>I have been in the same position as the original poster, >>and I certainly sympathize; but I suspect that not enough >>people would really want such a feature when getting it would >>mean more overhead and diverting deveopment resources from >>other projects. >> >>That leaves a few other possibilities, such as writing your >>own system service or logging facility and hooking it into >>the operating system. This is not a job for beginners, but >>it could be done. > > Back some time ago, I submitted a bit of code which was a template > for intercepting change-mode system services. It appeared on the > OpenVMS Freeware CD (I believe V3.0) as SSINT. It came about from > a discussion Glenn Everhart and I were having concerning some RMS > I/O performance enhancements. What the intercept was to do would > have increased I/O performance significantly and several test were > done prior to implementing the intercept to prove it. However, the > overhead of the system service intercept scheme was enough to weigh > down the performance gained by the code that was hooked into these > system services. > > Such a 'hack' could be added to monitor/track logical name trans- > lations and also DCL global symbol lookups but I would not recom- > mend that it be left in place on any production environment. The > pervasiveness of logicals in VMS would yield and enormous degrad- > ation of overall system performance with such a 'hack' in place. > Tracking DCL symbols could be isolated to a process and thus, the > entire system performance, in that aspect, would not be so impeded. > Having the ability to use something like this as a "once in a while" tool was the intent. In my case I have two clusters set up to be almost identical to one of our main production clusters. While the remediation people are working on getting the code fixed up, I'd like to be able to "clean up" the account environments. I can't do dozens of trial and errors because it would cut into their schedule drastically. A couple of "oops" would be OK.