Re: A Customer's Request For Open Source Software

["Robert I. Eachus" <rieachus@attbi.com>]

Larry Kilgallen wrote:
> In article <3F53B88E.7040405@attbi.com>, "Robert I. Eachus" <rieachus@attbi.com> writes:
> 
> 
>>As I said, most RDBMSs already have the ability to bypass the OS and 
>>access disk drives directly.
 >
> They do ?  On what operating systems ?

I guess we have a different idea of what bypassing the OS means. I was 
thinking of for example, Oracle on Solaris.  You can go through the file 
system, bypass the file system and have Oracle manage a partition (raw 
mode), or give Oracle full control of the DISK device or devices.  But 
even then you are right, the DBMS uses the standard device drivers.  The 
difference is that if Oracle controls the entire disk, it can take 
control of the head positioning and other driver level functions.

> I know compilers that read source that way, relying upon the OS paging
> in the data as appropriate.  But it is an individual performance/coding
> tradeoff, not applied so universally as in Multics.

Well, in Multics it is universal because there is no user level 
distinction between main memory and disk storage.  It is all memory, and 
once you have linked to a segment, there are no further OS level 
commands to read and write it.  Well, technically not. You could break 
the link to the segment if you wanted, or were running out of segment 
descriptors in your process.  But breaking links was more likely to be a 
side effect of deleting the segment.

A bit of magic here.  If you were compiling a file that was part of an 
application the compiler would normally rename the old version of the 
code segment, not delete it.  That way if someone else--or you in a 
different process--were executing the old code, the segment was still 
there.  But you could then delete the old version to force users to use 
the new copy.  What if the segment you were compiling was part of the 
operating system?  Why should that be any different?  So you could 
modify Multics on the fly.

Occasionally with early versions of Multics, there were "flag days". 
There was an incompatable change to the OS, so at a particular point all 
the affected old segments were replaced with new versions.  But they 
soon stopped.  The reason that flag days happened was usually that 
someone had concluded that some data structure did not need an internal 
version number.  Pretty soon it became accepted wisdom on Multics that 
every OS data structure needed a version number, including structures 
that contained structures with version numbers.

-- 
                                          Robert I. Eachus

"As far as I'm concerned, war always means failure." -- Jacques Chirac, 
President of France
"As far as France is concerned, you're right." -- Rush Limbaugh