[Info-vax] HP wins Oracle Itanium case

[John Wallace]

On Aug 24, 2:38 pm, koeh... at eisner.nospam.encompasserve.org (Bob
Koehler) wrote:
> In article <gmtZr.3213$_h6.... at fx03.am4>, ChrisQ <m... at devnull.com> writes:
>
> > I'm not sure that logic would make so much sense now...
>
>    My first UNIX did not have shareable libraries.  So every program
>    that called printf has a copy of it, ...
>
>    MS-DOS was like that, too.  Hello World in Fortran was huge on disk.
>
>    But UNIX and Windows caught up with VMS on that one a long time ago.
>    You need record support for several HLL, but there only needs to be
>    one copy of the code and read-only data.
>
>    What you don't get that way is the code you need to pass data files
>    between record oriented HLL and byte stream oriented HLL.



"[Windows] there only needs to be one copy of the code and read-only
data."

(sorry if quoting out of context)


Does the term "DLL Hell" mean anything to readers?

Or the alleged "fix" for DLL Hell, "side by side" configurations (aka
WinSxS, a term which readers can search for)?

Readers who understand these things will know that Windows [still]
needs multiple copies of the "shareable" stuff, a problem which afaik
the PDP11 OSes fixed back in the days of IAS and M+, a problem which
was fixed by design in VMS with its sensible architecture of shared
libraries from the first day I saw a VMS box (and probably long
before).

Readers with access to XP may wish to look in [C:\ or wherever]\Windows
\WinSxS\ (not sure of the details in later versions).

The first "modern" approach to fix this challenge under Windows was
that each application supplies its own private copy of the required
DLLs. That first attempt has since been replaced by a different
technique which also results in each app supplying its own copy of the
shareables.

So my home XP laptop with a relatively small number of installed apps
has for example 20 copies of the Visual C runtime(s), half a dozen
copies of some XML library, and so on.

I don't know what happens on these versions of Windows if an error is
discovered which requires a fix to the various copies of the VC
runtime (or the XML library, or whatever).


I believe that VMS has a mechanism for addressing that requirement.
Architected in from a long long time before NT was around.

For the compiler/linker aficionados: how easy is it to fix this with
ELF format objects and executables ? On VMS or elsewhere? Presumably
it's fixed somehow in ELF on VMS on IA64, or is that a risky
assumption?

I also believe that the number of people who care about this
difference is small, and the number of those who care who are in
positions of influence is even smaller.

In due course MS or Apple will re-invent the concept of properly
architected sharable libraries and it will suddenly be both visible
and important to the world at large. The next change of chip
architecture would be an obvious opportunity for such a change, but I
don't suppose MS fixed it in Windows on ARM.