Microsoft-Paul Allen

[Liam Proven]

On Sat, 20 Oct 2018 at 19:31, Tomasz Rola via cctalk
<cctalk at classiccmp.org> wrote:

> Oooh. My personal recollection about w95 is that there was a lot of
> touting before the premiere day, how advanced it was because "object
> oriented operating system". The premiere came, the toutings quickly
> faded away, never heard any kind of objection about this aspect. I,
> for quite long time, had been thinking W95 was a scam because for the
> life of me I could not spot any sign of its object-orientedness (and
> there was nothing else interesting enough to make me want to tinker
> with this... something).

I think the explanation for that is fairly clearly there in the history.

NT 3.1 came soon after Windows 3.

After the OS/2 "divorce" from IBM, MS ran its 2 big OS projects,
Chicago and Cairo, more or less in parallel.

Cairo  was next-gen NT, Chicago was next-gen Windows 3, at that point
intended to be "Windows 4".

Cairo started to fall behind schedule very early.

So more effort was given over to Chicago. A fair bit of the ambitious
UI work for Cairo made its way over to Chicago.

Cairo was intended to be semi "object oriented", with a
database-oriented filesystem (something Be did better in BeOS' BFS).
That never happened, but the object-based (rather than folder- or
drive-based) system browser made it over to Chicago.

The Explorer, as it came to be known, uses several "virtual folders"
-- "My Computer", "Network Neighbourhood". "Control Panel" etc. These
have no location in the filesystem, you can't manually put anything in
them or delete anything -- they only appear in Explorer, automatically
populated with stuff _drawn_ from the filesystem or the Registry.

Those are the vestiges of the Cairo object system.

In itself, these things are vestigial remainder of concepts in
NeXTstep, Xerox Smalltalk, HP NewWave and so on. By this stage, the
real meaning has been forgotten, and "object oriented" has become a
buzzword meaning, vaguely, that the user manipulates "objects" which
may not genuinely exist as files or folders in the filesystem. They're
virtual entities, generated by the OS on the fly.

> It was only years later that it finally came
> to me: I might have been one of the very few people who not only
> understood some of the buzzwords but also was duped into believing
> there should be some substance behind them (which maybe makes me
> exceptional, just not in a good way).

There was substance behind them once.

But, in a pattern that is very typical of the development of the
digital computer, especially microcomputers, the evolution goes like
this:

[1] someone, probably an academic, invents a new concept

[2] someone else tries to implement it, finds it hard, and has to
bodge it in some way -- with hardware extensions, or an abstraction
layer, or faking it up and presenting it as if it were real

[3] (a) another company copies the general idea but, lacking the
conceptual underpinning, simplifies it into near-meaninglessness

... or...

[3] (b) the other company finds a much quicker, simpler way to
implement it, such as by doing it in cheap software rather than
expensive hardware, or by some clever hack to another part of the
system.

Examples to illustrate my point:

[a] Microsoft decided to add an RDBMS to its new OS. (It's not
integrated from the start, like in Pick.)

[b] It talks widely about some of the things this will enable, such as
querying the filesystem like querying a database rather than
iteratively searching

[c] Be builds a new OS from scratch, and free from legacy
compatibility restrictions, designs a filesystem with extensible,
queryable attributes, thus achieving MS' goal with no database
involved.

[d] Apple fakes the end result of this by hacking a
file-modification-watching daemon into its Unix, enabling the daemon
to maintain an index for the whole filesystem. That in turn enables
near-instant searching, without needing a whole new filesystem.

[e] Microsoft having now been comprehensively outdone, abandons its
database-in-the-filesystem idea and tries to bolt-on a filesystem
indexer -- but because its OS is far more widely-used by a far broader
range of hardware and software, it can't do the low-level hackery
necessary without breaking legions of 3rd party apps, so the MS
implementation is poor and takes years & multiple product generations
to get working.

It's a sort of horrible sequel (see what I did there?) to the Osborne Effect.

https://en.wikipedia.org/wiki/Osborne_effect


> Nowadays, I consider W95 as very interesting subject of study - a
> technical product of non-technical genius(es) (ok, if there were tech
> geniuses involved in its making, I would say it does not show up).

I strongly disagree. (And I am no Microsoft apologist!)

On the one hand, the cosmetics. *Every* Unix desktop out there draws
on Win95. The Win95 Explorer re-wrote the book on OS UI design. The
_only_ company to resist was Apple, because of course, some of the
reasons that Win95 is the way it is are attempts to do things
differently from Apple so as not to get sued.

Look at other contemporaries of the original MacOS -- DR GEM,
especially Atari ST GEM, and Amiga OS -- and you'll see how they very
clearly draw from MacOS. And DR did get sued and PC GEM was crippled
as a result.

MS, a few orders of magnitude bigger, had to avoid this.

Secondly, the tech.

Look at Raymond Chen's "old new thing" blog and other sources for the
tech brilliance that led to Windows 3. Consider how very compatible 95
was with DOS drivers and Win3 apps. That was some virtuoso stuff. OS/2
(especially 2.x but also later), by comparison, was vastly less
compatible and vastly less stable, with big performance-sapping chunks
of 16-bit code, including the filesystem -- which even Windows for
Workgroups 3.11 replaced! -- it single input queue and so on.

OS/2 was and is a pig to install, to maintain, to update, it wasn't
very compatible with anything else, was totally incompatible with DOS
and Windows drivers, and wasn't all that stable.  And its UI was ugly
and clunky, IMHO.

Win95 was astonishingly compatible, both with DOS drivers and apps,
and with Windows 3 drivers, and yet it was 32-bit in the important
places, delivered true preemptive multitasking, reasonably fast
virtual memory, integrated networking, integrated Internet access, and
a radical, clean, elegant UI which *every single OS after it* has
copied.

> It
> took a lot of manipulation and wind sniffing to make it such a big
> success, and plenty of intellectual indolence from rivals and
> customers.

Not really, no!

There honestly wasn't anything to compare or compete with it.

Linux was very primitive and clunky back then.

OS/2 wasn't much better.

NT took years to catch up, only really getting there in 2000 and still
needing massively more hardware.

95 was an industry-transforming product, bringing much of the
user-level power and elegance of the Mac to the COTS PC platform.

Consider the anecdote about the swan:

http://www.horebinternational.com/the-swan-metaphor/

"BE LIKE A SWAN GRACEFUL BUT FURIOUSLY PADDLING BELOW THE SURFACE"

Considering the layers of hacks below the surface, Win9x was amazingly
stable and reliable, and used by more people than all iterations of
all other GUI OSes of the times up to then put together. It spawned
the entire industry of media-capable Internet-connected
home/entertainment computers.

I'm amazed as a confirmed MS skeptic that I have to defend it like
this, I must say!


-- 
Liam Proven - Profile: https://about.me/liamproven
Email: lproven at cix.co.uk - Google Mail/Hangouts/Plus: lproven at gmail.com
Twitter/Facebook/Flickr: lproven - Skype/LinkedIn: liamproven
UK: +44 7939-087884 - ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053