Frank Soltis, Father of AS/400, Retires from IBM

[efnkay]

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
** This thread discusses the Content article: Frank Soltis, Father of AS/400, Retires from IBM **
0

[cherev]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
I'd like to see an in-depth article (or two) by Soltis about how the AS/400 (iSeries, i5, System i - or whatever IBM is NOW calling the best platform on Earth this particular week...) OS can actually be improved beyond its base design. Prior OSes employed inarticulate semaphores in words or in matrices, whereas CPF and its heirs uses messages, which is far more flexible. What could be better than each job/thread having its own copies of code and temporary variables in working storage while operating on shared data? Anyway, in most business contexts many more processors just means more processors mostly waiting for something to do. The real choke point since the Nineties has been in getting data to the working storage directly accessible to the 'central' processors. What the AS/400 et al. could really use is redundant solid-state 'disk' at something near the price of rotating memory. Maybe with associated 'channel' processors that could execute logical-view or SQL (the real thing, not the M$-crippled crap) instructions. That way, 'central' processors and memory would only 'see' relevant data. BTW, the 'fresh boarder' classification imposed on Cherev is nonsense; I subscribed to Mid-Range Computing during the Eighties, have been getting the newsletters ever since MC went paperless, and have been writing letters (and e-mails) to the editors for nigh on 30 years.

[efnkay]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
Don't take offense...I always thought the "boarder" designation was referring to my age. Happy birthday. Hoo-hoo :-)

[H.Boldt]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
I know my opinions may be anathema to most of the visitors to this site, but here goes: We all know that the the S/38, AS/400, iSeries, etc. is nothing like most other systems out there. Some look at the unusual operating system and extol its advantages. But I wonder how popular the system could have been if it had a more conventional architecture? Consider the S/38. Those of us who worked with it knew it had performance issues. These were alleviated somewhat by coding conventions, but these conventions also made maintenance more difficult. Heavy memory and disk requirements made the system an expensive proposition. The 16-byte pointers forced unnatural coding. The MI architecture, while geared primarily to RPG, made compiler development more difficult for other languages. Some of the issues went away when ILE was introduced, and C became a viable coding proposition. However, the architecture still made porting code from other systems a non-trivial endeavor. Fast forward to today. Even lowly Intel and AMD CPU's have 64-bit addressing, and RPG programmers are still stuck with a 16M limit on data! Even S/370 took addressing "above the line" in the 1970's! Sure, there are some interesting aspects to the i architecture. But if the system had a more conventional architecture, I assert that it would have been cheaper and faster sooner, and there would be more 3rd party software ported to it from other systems. And in all likelihood, it would have had an appeal outside the SMB marketplace. Cheers! Hans

[cherev]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
The S/38 hardware was the problem, not the OS nor the paradigm. Internal IBM politics nearly sabotaged the mid-range offerings into death. The developers of mid-range gear had to fight the mainframe bigots tooth and nail to get any capability, because the S/370, /390, and 43xx crowd were determined to not allow anything to push performance from 'below'. But your post about the hardware limitations is stuck in the past, or is sly disinformation. The current address space for 'AS/400' gear is at least 64-bit. I dimly recall some bumphft about 128-bit words circa 2002...

[H.Boldt]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
cherev: You're right that the current power hardware has 64-bit addressing. Then why oh why do RPG programmers still complain about a 16M limitation on data spaces and static storage? But I disagree with you on your first point. When designing the O/S, they made certain design decisions that seem questionable in hindsight. For example, the high demands on RAM and disk storage can be traced back to the single level store. While academically an interesting concept, it also required that pointers be huge. Back when RAM was expensive, this meant unnatural coding conventions, which, as I mentioned before, made porting to and from the system problematical. And whereas all other O/S's put each process into a separate address space, the single address space meant an inordinate amount of effort into assuring that processes don't walk over each other. Then, consider the MI. Experience showed quite clearly that that was an unfortunate design decision too. The intent was to make it easier for the compiler writer. But while it made some things easier (but unfortunately, not all!) for the RPG compiler, it made other languages unacceptably slow. Fortunately, ILE came along which made it easier for other languages. But still, the other O/S features still make porting code not fun. Again, my point was that the system could have been more popular if it had a more conventional O/S architecture. But then again, you're probably right that internal politics made the i what it is today. Today, the computing world is dominated by two operating systems: Windows and the many varieties of Unix. i and z are the last relics of the traditional monolithic proprietary systems. i and z have their places. But the really interesting things are happening (and always have happened) in the Unix world.

[cherev]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
"But I disagree ... When designing the O/S, they made certain design decisions that seem questionable in hindsight. For example, the high demands on RAM and disk storage can be traced back to the single level store." I suppose the software designers expected the 18-month silicon cycle to rescue them. But clearly they didn't reckon with those clever mainframe obstructionists. Had 32-bit memory addressing been implemented with the 'Hundred' series, along with better disk subsystems (IBM would have done better to early-on buy commodity SCSI disk arrays instead of using the 33xx junk), the single-level store idea would have even more brilliant. "And whereas all other O/S's put each process into a separate address space, the single address space meant an inordinate amount of effort into assuring that processes don't walk over each other. ... The intent was to make it easier for the compiler writer. But while it made some things easier ... for the RPG compiler, it made other languages unacceptably slow. Fortunately, ILE came along which made it easier for other languages. But still, the other O/S features still make porting code not fun." I don't recall any memory overlap problems. Of course, I never tried to code in MI either. Business problems mostly involve making the code and data match the real world, and then in controlling change. Using the highest level language available. Porting applications that don't tightly bind to strong data descriptors may be a mistake in itself. Might be better to totally redesign and recode such. "Today, the computing world is dominated by two operating systems: Windows and the many varieties of Unix. i and z are the last relics of the traditional monolithic proprietary systems. i and z have their places. But the really interesting things are happening (and always have happened) in the Unix world." Well, actually, Linux and the BSD Unix heirs are the areas in play. Vista and Win/7 are just more Barbara Streisand, Piled Higher and Deeper. I'd like to see IBM take a Debian fork, clean it up to the tightest NSA standard, and then sell the hell out of it as a Sys/i guest Web server. I'm some nervous about the Ubuntu distro invading offices, however.

[H.Boldt]

Re:Frank Soltis, Father of AS/400, Retires from IBM

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
"Well, actually, Linux and the BSD Unix heirs are the areas in play. Vista and Win/7 are just more Barbara Streisand, Piled Higher and Deeper. I'd like to see IBM take a Debian fork, clean it up to the tightest NSA standard, and then sell the hell out of it as a Sys/i guest Web server." The sooner we can push Windows aside, the better, IMO! If you want a secure Unix server, look at openBSD. Perhaps IBM should try to get involved with that project to push some improvements there. openBSD is incredibly anal regarding security, but it also means they haven't kept up with certain technologies, such as Apache2. Here, we like having openBSD on the perimeter of our web application, but part of the app requires Apache2! If openBSD supported Apache2, we could eliminate one whole tier. I don't think forking is the answer. We have more than enough choice in the Unix world as it is! I believe IBM is taking the right approach already - contribute to existing Linux distros. "I'm some nervous about the Ubuntu distro invading offices, however." The sooner we can push Windows aside, the better, IMO!

[cherev]

OpenBSD / Apache 2.x problem

** This thread discusses the article: Frank Soltis, Father of AS/400, Retires from IBM **
Have you figured out your problem yet, H.Boldt ? The problem probably isn't BSD nor any Linux distro, but the 'radical' changes to Apache v2. See: http://www.linuxquestions.org/questi...stable-452109/ Why does anyone need Apache2, except in an initial bare-silicon install situation? If the latter, try the Winmin tool. See as a place to start: http://www.webmin.com/cgi-bin/search...?search=Apache