🚀 go-pugleaf

In article <4244ro$bje@barnacle.iol.ie> vals@iol.ie (V. Sliouniaev) writes:
   fat@cnit.nsk.su (Dmitry V. Irtegov) wrote:

   > Remember, Win95 (aka Chikago) was planned as quick and dirty hack
   > implemented in one year and having short lifespan. In '92 it was 
   > planned to release in '93 and die in '96, releasing marketplace 
   > for NT and/or Cairo. The main idea was to implement new API with
   > minimal architecture change, kinda of Win32s.DLL on steroids.

   Along with putting a simple semaphore for non-reentrant system calls,
   they've delevoped and built in also:
      - 32-bit parts of GUI, 32-bit font engine with anti-aliasing,

Anti-aliasing font engine is really cool, you're rigth here.
But it is add-on, not a change of kernel. I'd respect Microsoft more
if they added this feature to NT first.

      - Multiple input queues: a simple (almost) 16 -> 32-bit port of a
   single-threaded applicaion doesn't lock the whole system when the
   application calculates something. Impossible with OS/2.

Not completely true about OS/2. PMEMACS is a quick port of non-threaded
Unix program. It isn't well threaded: while emacs meditates with it's 
"Garbage collecting..." mantra, PMEMACS even couldn't redraw window properly.
But it doesn't stop other PM tasks from accepting input.

Well, some simple method of Win16->PM porting might lead to PM-hosing
programs. It means developers should choose other methods of porting :).

      - 32-bit installable file system with built-in disk compression,

??? You mean if I'd write 32-bit IFS driver for HPFS or ext2fs,
those file systems will get disk compression? 8-O.
Seriously, rudimentary form of IFS was present in DOS since 3.30.

      - 32-bit multiple network support including DHCP and WINS clients,

Our local Win95 user was forced to install DOS-hosted IPX stack and
Trumpet Winsock - he wasn't confident enough in those 32-bit networks :).

      - PnP which I saw working with loadable network drivers,
      - Remote setup and administration support,
      - National language support, with multiple keyboard layouts, code
   pages, locale dependent sorting, etc.

Ouch... Switching from Russian to Latin keyboard is braindead and not
customisable, codepages.... Win16 apps do not see "Russian" versions of
TrueType fonts supplied with the system, so our Win95 pioneer was forced
to install old TrueTypes from WfWG setup and use them.
Locale dependent sorting is ANSI C standard - big deal....

But when Win95 telnet doesn't support KOI8->CP1451(?) translation,
I'd say it is big flaw, because it makes that telnet 100% unusable for me.
In OS/2 I use C-Kermit that supports KOI-8 out of the box, and if it doesn't
support something, it has customisable translation tables. It is what I
call "National language support".

      - Help engine with full text search (works for old files like
   win31wh.hlp, too),

Help engine doesn't touch a way the kernel works.

      - OLE2, new shell, new look and feel (is it new though?)

32-bit OLE is implemented in Win32s 1.2.
Shell... well, shell is new, but if I'd install new version of bash on
my UnixWare box, will I have new OS?

      - WinG, 

Was available for Win 3.x for significant time.

      - ...

Well, only thing that doesn't match my opinion about Win95 as 
"Win32s on steroids" is anti-aliased font engine, because font engine
is simportant part of GDI. All other items either are application/C library
level things, like locale-dependent support, or not as new as you imply.

   > Later it appeared that it cannot be _quick_ hack, but it is still dirty :).

   ... but much more pleasant to look at than Win31 or OS/2.    :^)

Well, to make OS/2 look as pleasant as Win95 does, I need several mouse
clicks - change folders background color, icon and menu font to Helvetica,
and some other minor thing. I never seen Object Desktop however...

Anyway, I'd prefer solid OS with mediocre view to shit in nice shrinkwrap.

   > It is crippled for "compatability" reasons, but it isn't 100% compatible.

   99.9%. 

Looking on application glitches list, rather looks like 70% :).

   I think it's "crippled" mostly because of development cost
   reasons. It is compatible enough to meet their goals (not hackers').

It was exaclty my point: Win95 is crippled to carry as much legacy Win3.x
kernel code to new system without change, as possible. I do not know was
it done due to "development cost" (I'd rather cut on Rolling Stones song
and put these money to development ;) or due to strong lobby of Win3.x
developers who don't want to be replaced by Cutler's team.

   MS concentrated on more important issues. Even though the system runs
   viruses and Cyrillic font TSRs as well as Mac DOS emulator does. :)

I repeat again: MS concentrated on "Look and Feel" and advertisement
issues instead of kernel development. It remains to be seen which of
these issues are more important.

Actually, it is very interesting experiment: Would mass computer market
accept empty box with nice cower and enormous hype? Or it wouldn't?
I'd like to see results of this experiment.

BTW, it is very similar to western "left" intellectuals who called
Stalinism "interesting experiment" back in '30s. I'd like to see results,
but I don't want to be _subject_ of that experiment.

   > It uses old DOS configuration in order to simplify installation,

   I.e. setup is clever enough to look at your old good working DOS
                                                        ^^^^^^^
Key word. How setup might be sure your DOS configuration is working?
As far as I understand, many install horror stories are due to broken
DOS configuration.

   > but it's installation still isn't flawless, according to c.o.m.w.setup.

   It's an ideal.

You call not flawless instalation ideal? It might be "best what is possible",
but ideal?

[skip]
   [Would you go]
   through *all* the available manuals (because layout of information is
   bad,bad) and add some cryptic lines into config.sys, or would you
   change a value in a dialog box field, which will be checked after for
   possible invalid values? (If this doesn't help, one can use full text
   search into on-line docs).

Well, what if that nice dialog box doesn't inclide right value?
When I have human-readable configuration file, I have last resort for
the case automatic installer broke something. But with GUI registry,
if something goes seriously wrong, you are totaly lost.
IMHO, best way is to have human-readable text file with full documentation
for each statement and GUI (or text-mode menu) driven config utility.

   > From technical standpoint, Win95 is absolute failure: none of proclaimed 
   > goals were reached. It remains to be seen will hype compensate this or will 
   > Win95 fail on marketplace too.

	   It supports important parts of Win32 API, including multitasking and
   network support, well enough to be able to run almost all applications
   designed for WindowsNT (I mean, clients, not server apps).

It remains to be seen whenever these parts are _important_ or not.
Just one year ago most Microsoft advocates tell us they aren't :).

	   It fully supports Win16/DOS also (including Borland 3.1 Windows full
   screen debugger), which means that the users can use the system for
   their real work w/o either complicated/expensive changes in HW or SW
   upgrades.

First, not "fully", but "mostly". In practice "mostly" is often enough,
but sometimes it is serious practical difference :). It remains to be seen.
Also, upgrade from 4 to 8 Mb is rather expensive HW upgrade, and in
many cases you should went through motherboard upgrade too: many old
motherboards do not support 8Mb SIMM configurations.

	   API is fully documented and very well supported (MSDN Level 1,2), and
   runs on many platforms (as WinNT). Many compilers are available too.

WinNT will run VxDs? Non-Microsoft platform will provide Win32 compatibility?
I only heard of IBM plans to provide source-level compatibility in OS/2,
and it is 1. isolated case and 2. vapor.

	   IMHO, Win95 is not a failure -- this is, in fact, a very well designed
   and "thought out" transition system. It has its own limitations by
   design, both  for compatibility reasons and, well, cost of development
   too (why design processor-dependent-DOS emulator for the low-end
   system which will last only two years?).

See above on compatibility, development cost and all that.

Microsoft already has DOS emulator: NT DOS box. It'd be better to
enhance this emulator by compromising OS security (or providing an
option to compromise security, like it was done in OS/2) than to
start Win95 kludge project. 

   So, unlike Windows NT, Windows 95 can be used in place of Windows to
   run weird 16-bit applications (with their VxD, bad DOS calls, etc.).
   Also, unlike OS/2, it can run fully blown 32-bit applications too. 
   And yet it introduces some advanced features ahead of Windows NT.

First, OS/2 can run full-blown 32-bit PM applications, so I don't see
problem here.

Second, I'd respect Microsoft more if Win95 thing was based on NT kernel.
Stripped down NT with options to allow VxDs and direct hardware 
access for DOS box, DOS box idle detection and other OS/2 like stuff
and new look and feel would be killer OS. I do not beleive NT wouldn't
be stripped down to comfortably run in 8Mb, especially with same effort
that was put in Chikago project.

Yes, allowing VxDs and direct hardware access would compromise security.
But win95 has no security at all, it even doesn't protect it's own kernel
address space - and nobody seems to care, except "hackers". Having this
as an _option_ would be ideal - you might choose between security and 
compatibility. (You shouldn't have this dilemma at all, but with DOS
it is impossible).

But Microsoft choose to make cludge instead of compromisse OS.
I wonder why it happened:
1. People who make decisions in Microsoft are utterly incompetent.
2. They deliberately produce shitty products for some reason:
 2.1 because they are members of some conspiracy, probably including
     some extra-terrestrial influence, in order to slow down technical 
     progress of mankind :).
     Performance history of Microsoft doesn't contradict this idea:
     they always had good product and shitty product (Xenix and DOS,
     OS/2 and Windows, NT and Win95) and always chose shitty product
     over good one. Well, "good" product each time had it's flaws,
     and "shitty" product was somehow functional, but pattern is clean.

 2.2 because they want to experience their power ower the userbase,
     like in 1984: "if you don't make them suffer, you aren't sure
     they obey". Sorry for citate, I haven't read english text of the book.
      This version also doesn't contradict MSoft behavior:
      Why in the hell Microsoft made Russian encoding in Windows 
      different from Russian encoding in DOS, thus making any 
      DOS+windows system heterogeneous environment?
      Why in the hell Microsoft did Russian encoding in Windows
      different from any other widely accepted de-facto, governmental
      or international standard?
      Finally, why in the hell Microsoft patched GDI in russian version
      of Windows, so it uses other encoding of russian letters in fonts,
      so fonts supplied with third-party russian drivers aren't compatible
      with "true" Microsoft russification?

     Only intelligible reason for this: Microsoft wanted to see,
     will Russia upgrade to Windows despite all this shit or not.
     If it won't, it isn't that big share of world market.
     If it will, it means Microsoft would do with it's userbase everything
     it wants, they will upgrade anyway.
     Probably, Win95 thing is same experiment, but repeated on larger
     scale.

You are welcome to propose other variants.

   > Fatal Error: Couldn't enter Windows, use Doors.
   Some doors have "NO EXIT" sign above them. Think twice.

Worst case is Microsoft Windows: they have no exit nor warning sign.
You enter and couldn't go further, because there is no upgrade path
without significant pain.

Cheers,
	Fat Brother.

Fatal Error: Couldn't enter Windows, use Doors.