[Info-vax] Ignorance promenade: On the nature of patching VMS

[Marc Van Dyck]

Bob Gezelter has brought this to us :
>
> Subcommandante XDelta,
>
> Actually, the elimination of binary "patching" is not so much a switch in 
> architecture from VAX to Alpha, but the advent of compiled languages in 
> system-level products.
>
> When writing in traditional assembler, it is not difficult to determine 
> precise addresses to modify (frequently inserting a out-of-line transfer or 
> call if more space is needed). When using a compiled language, such 
> micro-level detail becomes difficult to manage for the purposes of patching. 
> It is far simpler to replace the component image with a new version.
>
> The kernel-level OpenVMS components are divided into "execlets", which are 
> individually loaded at boot time. Could one replace an execlet with a 
> modified version? Of course, the answer is yes, and replacing an execlet does 
> not require a global re-link (otherwise, software corrections on the level of 
> replacing execlets would be infeasible).
>
> A philosophical note: Studying the mechanisms used to maintain the system 
> will often illuminate ways in which the system can be modified. OpenVMS is 
> particularly consistent in this way, there are few, if indeed, any, special 
> gimmicks used for maintenance that are not utilities used in the normal 
> course.
>
> - Bob Gezelter, http://www.rlgsc.com

The necessity to use patches to provide system updates was also a
consequence of the limited size of the media used to distribute those
updates. With the advent of CDs and DVDs, it made no sense anymore.


-- 
Marc Van Dyck