Re: [Freedos-devel] FreeDOS limits! and FDNPKG v0.93a released

[Rugxulo <ru...@gm...>]

Hi,

On Wed, Apr 24, 2013 at 11:12 AM, Charles Belhumeur
<chb...@gm...> wrote:
>
> I'm new to the whole FreeDos intitiative.  If I might be allowed to add my
> two cents.  I'm working on an application for bioinformatics research.  The
> search for the best OS for this led me to FreeDos.  Although I did end up on
> picking Windows XP SP3 for the large user base and significant work to date
> debugging the OS.  I nixed FreeDos because the gene sequence files are too
> massive, some in the area of 300 GB.

XP's NTVDM works better (IIRC) than its successors. Despite some of
XP's bugs that were never fixed, DJGPP worked around 'em (for 2.03p2
and 2.04), so it works well there. Though I don't have any working XP
machines anymore, sadly. (DOSEMU and native FreeDOS have different
quirks and bugs, esp. regarding GNU AutoTools.)

Though that won't help you with big files. DOS itself usually only
supports 2 GB (e.g. FreeDOS) with some variants (e.g. Win9x)
supporting 4 GB files with weird hacks (that aren't universally
supported elsewhere). I guess you could (non-commercially) use EDR-DOS
with FAT+ (and manually call the API), but I've not tried.

I don't see how having one big 300 GB file is necessary. Just use 200
files of 1.5 GB each.   ;-)

> The problem with Windows and LINUX for science work is they're kind of like
> the family station wagon.   Trying to be all things to everyone.  Maybe even
> worse they my be an RV monster will all the amenities including the kitchen
> sink.  This makes these OSs too big, glitchy and they require far too much
> maintenance and support.  All of which gets in the way of scientific work
> and adds layers of needless difficulty for scientific workers.  The
> biologists who do grad studies in bioinformatics seem to have a bad time
> with all the arcana.  It stalls their careers and research for at least two
> years.  Cures for diseases and death for all us are similarly stalled by the
> overcomplicated IT.

Trying to keep backwards compatible while adding every new feature is
difficult. x86 has been around for 30 years, so that's a lot of cpus!
Even Windows 8 only runs with later-model P4s or newer. Everyone keeps
(re)inventing every tech in incompatible ways. (And keeping up with
constant kernel upgrades in *nix is annoying.)

> When I was a student we had these HP workstations based on Motorola chips.
> The ones I used had an OS based on Berkeley BASIC.  They were powerful,
> single tasking, big disks, lots of RAM in flat memory model, rock solid and
> spit simple to use.  In many ways ideal boxes for scientific and engineering
> work.  The Rocky Mountain BASIC compiler was spit easy for application
> developers.  No need to malloc, compartmentalize like C++ conventions and so
> on.  Just DIM and use the standard BASIC conventions.  I knew engineers who
> wrote a complete GUI under this scheme in a few months.

For whatever reason, developers think differently these days. They
overdo everything and use heavyweight solutions.

> I've often thought these modern Intel boxes had the potential to be similar
> platforms with the right OS.  That's why FreeDos caught my eye.  I think
> there's a niche with big potential for FreeDos in science and engineering
> work.  My advice would be, and I don't say this lightly since I know what's
> involved, bite the bullet, write whatever it takes to give users full access
> to all the RAM and disk space under a DOS style interface.   It only has to
> be done right once and it will open up a lot of potential on these boxes to
> a wide user base.

DJGPP v2 (DPMI) already is fairly well working with reasonably high
amounts of RAM (2 GB?). FreeDOS with FAT32 lets you use up to 2 TB of
disk space for a partition (though only max 2 GB file sizes). No real
PAE (yet) or 64-bit (ever) support, obviously.

Another good option for FreeDOS is OpenWatcom + HX.

So you can basically emulate subsets of POSIX or Win32 with these
toolsets, if native DOS (x86 assembly) or pure standard code (e.g. ISO
C, ISO Pascal) isn't to your liking.

> You may want to approach these people http://www.htbasic.com/ with the
> possibility of porting their compiler to FreeDos.  This is the modern
> version of the BASIC of the old HP workstations.

I looked at this yesterday. It seems to be a commercial clone of RMB.
Apparently they already had DOS ports, latest apparently being 6.0
from 1998. Unfortunately, I don't see any way to acquire it except by
buying either the legacy bundle or similarly using a newer product's
serial key to get a compatible older serial number and password for
it. And they are a bit pricey, to say the least. So I'm not sure how
useful this is for FreeDOS, "as is", but hey, at least it exists.

(The glib, over-simplified answer would be "Just use FBC" or similar,
but I suspect that's not as easy or reliable as it sounds, in such
cases.)

> The old HP workstations had an interesting feature in their graphics
> hardware.  They didn't have separate text and graphic modes. No need for
> switching.  You could dump graphics and text to the same screen in its
> single mode.  This was a very nice feature for app developers.  HTBAsic
> still supports this.

A lot of stuff is forced to GUI these days. You could argue it's for
technical reasons (fonts for Unicode), but a lot of people (MS?) also
think everything that isn't using a GUI is obsolete.

As mentioned in other mailing list threads, FreeDOS can use several
GUIs (mostly as third-party libraries) but doesn't require any by
default.

> Finally I'd like to comment on "The Wankers".  There's wankers all over
> science and engineering, especially in IT, who think making things
> complicated and arcane makes them look like geniuses. F**k No!  The whole
> thrust of science is to simplify complicated phenomena into principles
> everyone can understand and use.  I always point out Einstein to the
> wankers.  They hate that sh*t!  The UNIX/LINUX crowd seem especially prone
> to "wankeritus".  (F**k I really hate the wankers after all these decades in
> science and engineering!)

Managing complexity is not easy, but sadly most don't even try. They
don't minimize requirements, just stick to de facto standards that are
too brittle to reliably work (e.g. GNU AutoTools bugs, at least re:
DJGPP). That really bugs me, but it's too much complexity for me to
understand enough to (weakly try to) fix.

About Einstein and simplicity, he's often quoted by Niklaus Wirth in
his Oberon work. Though some people still use Modula-2 or Pascal or
some variants (or on *nix usually stick to C99 or C++03 or Obj. C v2
or ...). It's hard to get people to agree on anything:  compilers,
cpus, languages, libraries, build tools, etc.

> The whole FreeDos initiative seems to be a group of people who get
> this.  I can't tell you how refreshing and novel this is after 30 years
> of watching the simple basic functions of IT get twisted into the
> overcomplicated mess we now have.  Keep it up people!  Well done!
> Do whatever it takes to crack into a significant user base.  I think
> the time is right, the need is there.

I don't know. DJGPP would probably be the best place to enhance
things, but these days, volunteers are few. Again, most of it is just
ports from GNU, but their POSIX-heavy tools tend to not work well on
non-*nix (except sometimes with Cygwin or MinGW).

> (Anybody remember VAX-VMS?)

CWS (of CWSDPMI) always raved about it. In many ways, he helped make
DJGPP as "rock stable" as VAX/VMS was renowned for. Unfortunately, as
successful as DJGPP has been, these days it isn't appreciated as much
as it deserves. (Though DJ works for Red Hat and, for now, still
supports DJGPP on the side [website, ftp, mailing list] with the help
of a select few gurus.)