[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: OT about Asus, was Re: What is the point of RAID?



Ron Johnson wrote:
On 11/10/08 05:38, Napoleon wrote:
Ron Johnson wrote:
On 11/09/08 17:53, Napoleon wrote:
Ron Johnson wrote:
On 11/09/08 06:58, Mark Allums wrote:
Ron Johnson wrote:
On 11/08/08 23:25, Mark Allums wrote:
[snip]

But, would you want a render farm made up of SGI workstations from the
1990s?  The state of the art is still moving pretty fast.  Even for
mainframes, the shelf-life of what is generally considered useful for a
lot of applications is less than 6 years.

Unless:
(a) your workload growth is relatively static, or
(b) you purchased excess capacity and are growing into it.

An important reason, though, why many mainframe shops upgrade is that the cost of maintenance contracts skyrocket after 4ish years, so that it's cheaper to buy a new machine than to maintain the old one.


True. But the software running on the thing is 45 year-old-COBOL. :)

Possibly the same binaries!

In a similar vein: contrary to Unix lore, most C apps are horribly non-portable, whereas COBOL apps are *very* portable.


Contrary to popular lore, COBOL apps are typically no more portable than C apps.

For more than a decade, small-systems COBOL vendors have made compilers that compile and run VSII/CICS/DB2 apps on both *ix and Windows.


And for more than 40 years, mainframe COBOL vendors have made compliers that compile and run apps on DOS and MVS systems.

That does not make them portable.

Sure it does, if other vendors' compilers successfully compile the code on Windows and *ix.


That's a mighty big IF.

And you can also say the same for C, FORTRAN or any other compiled language.

ANSI C code can be portable at the source code level. So can ANSI COBOL or virtually any other code conforming to ANSI standards. But very little of it does because the ANSI standards are quite limiting. For that reason, most compilers have their own extensions to the language.

Binaries are NEVER portable across different platforms (CPU and OS family). Pseudo-code like that generated by the Java and Smalltalk compilers is can be run on any system with the proper interpreter.

But compiled languages must be compiled for each platform. And if you need to use different compilers, you need to stick to ANSI standards (and all of their limitations).

I presumed that you were bright enough to realize that *of course* I was referring to source code portability... Only a raving idiot would think that you could take an MVS/OS binary and run it on a Windows or *ix box on a completely different CPU architecture (Hercules emulator not withstanding).

Maybe I overestimated your intelligence.


I understood what you said. But you don't understand the just because a program is written in COBOL doesn't mean it's very transportable. That is patently NOT true. It wasn't true when I had to deal with COBOL over 30 years ago, and has only gotten worse.

I guess I overestimated YOUR intelligence.


Reply to: