On 6/15/2023 12:48 AM, Zane Healy via cctalk wrote:
On Jun 14, 2023, at 10:06 PM, Chuck Guzis via cctalk
<cctalk(a)classiccmp.org> wrote:
Well, I don't know about the 640KB quote, but
he did say that OS/2 (what
was to become Warp 3) had Microsoft's full commitment. (cf "The OS/2
handbook"). At the time he said it, MS was already developing NT and
had not the slightest intention of honoring that statement.
What I also remember is this:
For the developers who paid for the advance copy of the "new OS/2"
(about $3000, IIRC), MS sent out copies of NT 3.1. Under threat of
lawsuit, they relented and refunded the advance payments.
--Chuck
When shipped, Visual Basic v1 stated that it was for Windows and OS/2. I
purchased it due to this. They only included the Windows version.
By the time MS released Windows 95, I had a system running Lotus SmartSuite for OS/2,
which by then was owned by IBM. Lotus SmartSuite was so buggy that I saw the writing on
the wall, and bought my first Mac, a PowerBook 520c. I’m typing this email on a Mac. If
MS had delivered an OS/2 version of Visual Basic, there is a good chance I’d be typing
this on Windows.
Zane
I worked for 3 years learning the internal communications APIs on OS/2
to support and enhance a common networking layer for Compuware
applications called "Generalized Communications Services (GCS)" as part
of the "GCS" team at that firm. The layer was used by all of the OS/2
front ends connecting to legacy OS/370 mainframe apps that Compuware
sold (Xpediter, File-Aid, etc.). I used to joke that all of that skill
was then useless as the world moved to NT. In reality, there are lots
of similarities between the OS/2 and WinNT IPC and networking APIs, but
that made for a less interesting story! :-)
GCS would no doubt cause people here to cringe, but I think it was
impressive for the time. It provided a common naming service (ala DNS)
for protocols such as LU.2, LU6.2, Named Pipes, NetBIOS, and TCP/IP, and
offered a way to simulate full duplex operation over half duplex LU2
channels. It also worked with all of the Windows 3.1X 3270 emulators of
the day and offered the ability to send arbitrary length data over LU2,
even with the significant constraints of LU2 screen sizes. Efforts were
helped by realizing that most emulators would allow for very large
screen sizes (128x128, or sometimes even 128x512, which would allow 16kB
- 64kBto be transmitted in one packet). There was a "hub" defined which
could seamlessly convert from one protocol to another if the sending
code found out from the naming service that the receiver was on a
different protocol, as the sender would send to the hub, which would
convert and then send to the recipient.
The emergence of TCP/IP on Windows for Workgroups, Win95, WinNT3.1, and
zOS spelled the end of GCS.
As a further anecdote for those caring about 370, a man by the name of
Walter Falby at Compuware got so frustrated at IBM's initial TCP/IP
protocol stack for the zOS environment he borrowed by Comer networking
books and implemented (in 370 assembler) an entire TCP/IP protocol stack
(called HCI) I think Compuware sold to customers for a while until the
IBM code got better... That developer reminded me so much of the "Mel"
of Internet Folklore fame.
/me still has his copy of OS/2 v3 and Warp sitting here.
Jim
--
Jim Brain
brain(a)jbrain.com
www.jbrain.com