[fpc-other] [fpc-devel] OS/2 and DLLs

[Mark Morgan Lloyd]

Ralf Quint wrote:
> On 12/18/2014 2:09 PM, Mark Morgan Lloyd wrote:
>> I presume that the early enthusiasm had largely dissipated by the time 
>> IBM started pushing it as a 32-bit OS, with a new binary format etc.
>>
> OS/2 is 32bit (for the most part) since v2.0, released in 1992 and it 
> became actually really popular after that with OS/2 3.x and 4.x 
> ("Warp"), so I doubt that this is a valid conclusion.

I'm about to be quite negative about OS/2, but I'd be the last to say 
that nobody should use it or support tools for it.

By the time the 32-bit variant came out OS/2 was already dead. The 
initial flurry of interest- and I was selling development tools at the 
time- wasn't matched despite the technical improvements: comparatively 
few end-users wanted it, it had limited penetration into the 
small-developer industry, and people writing tools and- in particular- 
libraries were far more interested in supporting Windows and co-existing 
with Desqview than chasing something with highly uncertain prospects.

IBM didn't know what the hell to do with it. They tried to target end 
users by getting them to upgrade their Win-16 systems to OS/2, but this 
effort was plagued by bad testing, poor installation software, and 
hardware with subtle bugs which meant that many computers quite simply 
weren't up to it. They hardly helped themselves by doing utterly 
lackwitted things like setting the startup noise to full volume: that 
used to cause me extreme embarrassment when installing it in open-plan 
offices.

Networking was in there, but in some cases was so broken that it wasn't 
even possible to log into IBM's upgrade service to see if the fixes were 
still available. Network configuration was also well beyond what a 
non-specialist was capable of, and support software would fail with 
meaningless (to the typical user) messages about containers etc.

It's reputed that one of the IBMers responsible had previously attempted 
to steer Borland towards corporate customers rather than sticking to its 
traditional base of small-to-medium scale developers, and that she 
learnt her craft at CA who had followed the same path. I have no 
first-hand knowledge which confirms or denies that.

The bottom line is that from the early 90s onwards, the only thing that 
kept OS/2 going was corporate "seats", i.e. the number of office systems 
which used OS/2 as a specialist frontend to IBM mainframes. Full 
recognition of the necessity for a 32-bit OS was at least five years 
away, in the same way that the industry didn't start adopting 64-bit 
hardware for at least ten years after DEC introduced it.

I probably kept on using OS/2 longer than most, because (a) I was 
postprocessing NE-format files into a format for an embedded bare-metal 
'386, and (b) it allowed sufficient access to e.g. the floppy controller 
that I could test stuff on a development machine that would have been 
impossible with NT. But it was painful, particularly after it became 
obvious that at least some versions had timing loops that precluded them 
from being run on newer hardware. And that's why I've got a PC-310 under 
my desk.

The bottom line? If somebody wants to use OS/2 it's his choice, and if 
somebody wants to keep a development tool going for it I believe the 
community is enriched by his effort. Now would somebody please take the 
copies of Communications Manager and Database Manager that IBM told us 
we needed off our hands.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]