Hi,

On Tue, Apr 22, 2014 at 1:32 AM, dos386 <[email protected]> wrote:
>
>> I personally don't care what size the exes are

But he followed it up with "just as long as I can at least fit a few
critical things on a floppy." !!!

So that's far from indifferent! (But I personally think such
sentiments, while reasonable, are a lost cause. Almost.)

> So do many others. Result: Wing-ZIP 100 MiB ...

WinRAR 5.10 beta is less than 2 MB.

> Abobe-ACRO 100 MiB ...

MuPDF is a lot smaller, the 1.4 version's .ZIP is "only" 9 MB.

> "Hello world" 30 Byte's to 10 MiB (depends from compiler) ...

* disable code/data alignment (x86 doesn't need it)
* optimize for size, not speed
* strip debug info (etc.)
* keep add-ons as optional downloads
* use a smartlinker (--gc-sections, /OPT:REF, -CX -XX)
* compress with UPX (or similar)
* don't support bloated extra features (if you don't need it) just to be trendy

Though a lot of this is just bad (bloated) code generation from
compilers. Use a better compiler (PowerBASIC? FreePascal?). Note that
relying on MSVCR*.DLL is not worth the tiny size savings!

Whatever happened to the idea to "do one thing and do it well"?? It's
when you try to do everything that you can overburden yourself. But of
course customers demand (more) "features features features"! Obviously
I'm not opposed to features, but sometimes minimalism really is
better.

> Linux 1 to 20 GiB (depends from distro)

In fairness, yes, almost all modern distros need about 2 GB (minimum)
these days, even FreeBSD. Though "Linux From Scratch" claims to be
able to make an Apache-only mini-distro in 8 MB.

The problem with Linux (and DOS) is too much fragmentation. Too many
choices, too many optional pieces. (Actually, Windows is bad about
this nowadays as well.) Okay, maybe "problem" isn't the right word,
but it often makes people stumble. Some people just want everything
and the kitchen sink all bundled together, and they want it to "just
work" without lots of fiddling, downloading, etc. But that's not good
for slow connections or legacy hardware, so you have to prune things
and split them off sometimes. Unfortunately, that makes things harder
to keep track of! And yes, compatibility is also both a blessing and a
curse (though I'm partial, obviously). The less you have to do, the
easier it is to succeed. But one size does not fit all.

>> anyone noticed HXDOS?
>
> YES. Noticed 10 years ago, development stopped 5 years ago.

In fairness, it's quite stable and works fairly well, all things
considered. I think most of his recent work has gone into JWasm (et
al.) and trying to use those to (re)build HX.

BTW, it's impossible for HX to support everything. And if it's not
good enough by now (to some people), they'll probably never be happy.

>> How long has it been since you've done performance tests on Java?
>> JRE 1.5 was a HUGE performance jump from 1.4, and 1.6 was another
>> smaller but still significant jump from 1.5.
>
> The start-up takes centuries, and eats away several GiB's of RAM.
> Maybe your test program runs indeed faster for you ... supposing that
> that whole thing doesn't crash due to lack of RAM even before. And
> I haven't seen (usable) JAVA for DOS yet.

There was a commercial Java (1.1? "JavaPC"?) back in 1997 built with
DJGPP! (Though I never used it.) I have no idea what happened to it.
And modern developers are so annoying that most code probably requires
1.5 or newer (or even only latest / greatest, ugh). So it's probably
not too useful for us, even if we could dig it up.

You could probably use a third-party minimal Java clone to get halfway
there, but I'm not sure it's worth it. At least, from what I can tell,
people don't want a minimal console-only language and libs. You know,
ANSI C or ISO Pascal is great, but people don't want that simplicity,
they want all the non-portable, bloated, complicated, modern stuff. So
if you don't have GUI, HD, OpenGL, multi-touch, networking,
multithreading, Unicode, 64-bit, garbage collection, tons of libs,
OOP, generics, exceptions, etc. etc., well then they don't want it!
Nobody wants "minimal"!

Even GCJ is not widely used where supported (but doesn't work with
DJGPP anyways). At least I'm not aware of a lot of uses. AFAIK, most
Linux stuff (70%?) only uses C/C++. In contrast, C# (managed, using
JIT) seems pretty popular on Windows. BTW, I'm obviously not saying
that nobody uses Java, just that it isn't that widely prevalent on
Linux. Presumably big corporate types / enterprise with Oracle
contracts prefer it more than home users. (Yes, I'm vaguely aware of
IcedTea and Red Hat's fondness for it.)

>> > I'm thinking more and more there's no big niche for FreeDos.
>> The niche is running legacy software from ye olde days (or similar)
>
> That niche is filled by DOG-BOX/EMU, but maybe there are other niches left?

DOS on modern IBM PC clones have too many hardware compatibility
issues:  power management, lack of networking (almost no packet
drivers), no soundcard drivers (or even static libs), almost no USB
support. It doesn't look like most hardware companies care enough to
waste time on it. Heck, half the time they don't seem to even properly
support Linux. And DJGPP isn't exactly brimming with enthusiastic new
users or tons of ports from Linux maintainers either.

So expecting anything beyond what we've already got is probably naive
(sadly). They probably (rightly? hope not!) think that DOS will really
disappear once the BIOS is totally replaced by UEFI in all new OEM
shipments. I know some people think some partial BIOS compatibility
will be available, but I'm very skeptical.

In other words, it's complicated!

------------------------------------------------------------------------------
Start Your Social Network Today - Download eXo Platform
Build your Enterprise Intranet with eXo Platform Software
Java Based Open Source Intranet - Social, Extensible, Cloud Ready
Get Started Now And Turn Your Intranet Into A Collaboration Platform
http://p.sf.net/sfu/ExoPlatform
_______________________________________________
Freedos-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-devel

Reply via email to