Re: IBM i "Open I/O" Architecture

On 21-Oct-09, at 1:00 PM, midrange-l-request@xxxxxxxxxxxx wrote:

Thanks for your perspective, Jon. When Ian Jarman introduced Open I/ O, did you get the impression that it was only an RPG language feature, or that IBM would be opening up its I/O architecture?

This is _just_ RPG. Is there anything else?

While an RPG program may have no idea beyond the display file I/O buffer, the system-level routines that are subsequently evoked by display file writes and reads, do. If Open I/O is more of an architecture than a language feature, then extended display file meta data could be included in the Open I/O interface - just a speculative thought.

Interesting thought but as far as I know not part of current thinking.

In an earlier comment I inquired about whether the use of Open I/O might source code changes, and some speculated that it would - that it's intended for new types of applications, rather than a next generation of screen scraping - which may be fine. Personally, I prefer the idea of implementing Open I/O through additional keywords on the "F" spec, rather than something like an external trigger, because that would give the programmer more flexibility in using it for specific purposes, rather than globally - like triggers.

I think it likely that additions parms /F-spec keywords will involved. Not an external trigger and not like the Webfacing API where it is too late to do much. Extra keywords/parms would enable call-back into the code or other procs which would not work with a trigger type approach.

I like to think that Open I/O is more of an architecture, and that IBM will be opening it's remarkably streamlined interface for more purposes than the traditional ones.

It is a nice thought - but we need something that can be implemented by Toronto alone in as short a timeframe as possible. Full-blown architecture changes with Rochester-et-al involved just couldn't happen fast enough.


Jon Paris

www.Partner400.com
www.SystemiDeveloper.com