Re: Saving the System i: Fight Rather Than Switch

["James H H Lampert" <jamesl@xxxxxxxxxxx>]

"Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx> wrote:

> It's impossible to create a good-looking GUI using the 
> block-mode interface of the 5250. . . . It's your
> comparison of an event-driven, keystroke oriented user
> interface with the block mode architecture of 5250 that
> is flawed at the very lowest level.


Joe, you're better than this. You've got a more than 
adequate supply of functioning little grey cells between 
your ears.

Admittedly, as I re-read the paragraph, it's not as bad as 
it looked the first time, when I thought you'd said that 
the 5250 data stream itself was "flawed at the very lowest 
level." But still, saying that block mode is fundamentally 
incompatible with a GUI is something I'd expect out of 
somebody who spent his entire career working on 
asynchronous ASCII terminals (probably in remote-echo 
mode!) and single-user systems with integrated display and 
keyboard, where the only way to give users any kind of 
reasonable response time (even on the simple act of 
typing!) is with sub-second timeslicing.

The whole point of the 3270 and 5250 data streams (or of 
using large front-end processors to drive other data 
streams, instead of forcing the CPU to deal with them 
directly) is to off-load the business of keeping the user 
happy onto the terminal and its controller. And indeed, 
offloading the routine tasks of running a full GUI onto 
the terminal is not that much more advanced than doing so 
for a text-based user interface. Indeed, it's quite 
common: HTML is itself a protocol that offloads everything 
from keystroke reading to keyboard echoing to mouse 
gesture management onto the client.

Let's turn this around. For some reason, people have taken 
to calling 5250-data-stream terminals "dumb," but there's 
really nothing the least bit dumb about them; they're 
actually quite sophisticated, as terminals go. Let's talk 
about terminals that can TRULY be called dumb: so-called 
"glass teletypes" like the Lear-Siegler ADM-3 (as well as 
a number of similar units from Teleray). Indeed, for a 
time, LSI attempted to use "Dumb Terminal" as a 
quasi-trademark for the ADM-3. All a true dumb terminal 
does is scrolling 80x24 monochrome text. It doesn't even 
have cursor addressing: the cursor is captive to the 
bottom line of the screen. If you had an extensive 
background in 3270 and/or 5250 (even one limited to the 
most primitive forms thereof), and the only character-mode 
asynchronous ASCII terminal you ever laid eyes on was an 
ADM-3, would you even remotely postulate that a full-GUI 
X-Terminal, or a full GUI on a single-user system's 
built-in display, as we know such things today, would be 
even remotely possible? I daresay you'd laugh in the face 
of anybody who even suggested it!

An adequate GUI would be possible with 5250 terminals and 
controllers that exist today; all that's missing are the 
development tools and documentation. A really good GUI 
would simply require one or two more generations of 
terminals, with the 5250 Data Stream evolving with each 
generation of terminals, at about the rate it was already 
evolving as we went from 5251s to 3489s.

--
James H. H. Lampert