Gruess Gott,

On Tue, Mar 26, 2019 at 1:20 PM Tom Ehlert <[email protected]> wrote:
>
> > Developers are very sloppy and include lots of things that they don't
> > need. They also split up things into too many files. Too many
> > dependencies. I'm just saying, it's overwhelming, even for them.

I'm aware that this is freedos-devel and that perhaps you
misunderstood that I was specifically talking negatively about FreeDOS
developers. In fact, i was moreso (harmlessly) referring to bloated
GNU sources (that have lots of files for other OSes that we don't
need). The list of "minimal" GNU tools recommended for rebuilding
things is often ignored, and it basically assumes a mostly working
POSIX environment, which is beyond what normal DOS (or FreeDOS or even
DJGPP) requires.

* https://www.gnu.org/software/make/manual/html_node/Utilities-in-Makefiles.html

Perhaps you wrote this email without seeing my followup, where I
specifically was thinking of FPC and DJGPP, which are two of the
biggest offenders here (and a minor point of contention for Jerome). I
adore both projects, but they do not try for minimalism at all.
BinUtils alone takes up a ton of space, and both DJGPP and FPC require
it. So does FBC. (Actually, FPC doesn't forcibly require AS or LD
anymore except for startup code, IIRC, but it still includes them
anyways, for optional use.)

> for a clueless moron, you are making fairly big claims.

I never claimed to be a genius nor super useful. I only do what minor
things I can. I have to be humble here because I'm nothing special.
However, it seems unnecessarily rude to call me a "clueless moron",
even if you think it's partially true. I'm aware that those of high
intelligence have little patience for fools, but you always seem to
overreact (or else it gets lost in translation). Certainly someone who
is bilingual (or more?) like yourself should know that certain phrases
lack tact. Again, I'm no genius, but I have a large enough vocabulary
that I can imagine dozens of ways (both better and worse) to phrase
things. Remember, we're on the same team here, loosely speaking, even
if we don't really directly collaborate. Jim certainly wants me to
keep the peace here (but he's busy with real life lately, as always,
like most successful people). I really don't know why you're so prone
to grumpy outbursts.

Again, this was not a personal attack, but that should've been obvious
anyways. Not everything is optimal by default. It can be improved.
Yes, developers are *very* sloppy in licensing and dependencies and
rebuilding. It's a fairly persistent problem, especially for "niche"
platforms where there are few active maintainers.

> do you have ANYTHING to show to support these claims?

FPC.ZIP is 39 MB. At one time (a year ago?) we were using an old
(2011?) package that I personally made. I never updated it, and I even
suggested deleting it. I also didn't think any user directly needed or
used it. I figured people could download newest directly from upstream
and manually install it themselves. Then again, as already mentioned,
most people are assumed to have fast MegaBit Internet these days, so
it's maybe less important to worry over a measly 39 MB. Again, in my
second email, which you apparently didn't read (yet?) when you
responded to my first, I mentioned the possibility of an FPCLITE.ZIP
(which has been discussed at least once before, briefly, maybe
privately only to Jerome).

Even our package for Perl 5.8.8 is ancient (but best we've got) and 27
MB (half of which is bloated sources). I don't recall how easy it is
to rebuild but probably not very. It too probably requires a GNU POSIX
environment and who knows what else (NTVDM LFNs). A lot of GNU stuff
won't even build natively in DOS but instead requires Bash (only atop
NTVDM) or even cross-compilation.

For PSR Invaders, I used Sed (which is fairly minimal, at least
MiniSed or Cheap Sed) instead of relying on Perl + Devore's NoMySo
converter script. It would be horribly overkill and heavy complication
to rely on such a bloated, abandoned util. Even if Perl is useful
(which I presume it is, even though I don't grok it), relying on it
when a simple Sed script would suffice is what I call too heavy of a
dependency. The (old) package for GNU Sed that we have is "only" a
little over 2 MB. (Cheap Sed would be significantly smaller, less than
half a MB. But I've not made a package for that yet. There's too many
things to do.)

There are also other external projects (that I was thinking of, not
FreeDOS related) that require both Perl and Python. Or projects
(Watcom before becoming "Open" ??) that required four different
assemblers to rebuild (probably before they wrote WASM). Watcom
were/are very diligent about bootstrapping and relying mostly on their
own tools. Certainly you shouldn't need four different tools that all
do roughly the same thing. That's a big "no no" and something to
avoid, especially for (non-standard) tools that can't be publicly
redistributed freely.

> > Things need to be minimalized, simplified! But it takes a lot of time
> > and effort, and most don't care enough. (I know that's a cheap thing
> > to say, but trust me, it's a mess that should be avoided.)
>
> One of the rare times where you are right. It's a cheap thing to say.

People don't care enough to waste time on it. The IA16 toolset uses
LZMA compression inside the .ZIP to save space. But most of these
other packages don't. It requires more RAM to unpack, and even Mateusz
was worried about that (overall, not specifically here). Yes, we have
to sometimes think of such old machines where it will hurt. The
classic Info-Zip Unzip (6.x) doesn't even support LZMA (only BZip2).
So you have to use p7zip (which is flawed, see below). Well, or use
fdnpkg itself, which is presumably the best choice here. My point is
that it's confusing and does indeed require a slightly higher caliber
of machine than normally necessary, unless you want to just live with
bloat.

> >  Most things can be simplified and slimmed greatly without losing any 
> > functionality (without dirty tricks).
>
> just show your resume with this respect. else shut up.

I'm not bragging, and certainly I can't fix everything. I was just
trying to make a generalization and say that, in my experience, it can
be greatly improved if someone were to spend time on it.

I also mentioned my one-floppy (old) DJGPP setup again. That is
certainly smaller than 12 + 5 + 1.5 + .3 MB (total 19), which is
roughly bare minimum (although newer and "better" version) of what we
have now. The problem is that there is no perfect version. Too
bloated, too slow, too RAM greedy, too old, too weak, different
(newer) standards, and other tradeoffs make it unclear. Again, maybe
there should "also" be a -LITE.ZIP version same as how we have both
p7zip and (minimal) 7zdec (which I did indeed compile and package for
us as a simpler alternative since even my other builds of p7zip are
semi-functional but still quite heavily flawed due to upstream code
not directly caring about DJGPP quirks). Why rely on full p7zip (7.2
MB) when all you need is simpler 7zdec (105 kb)?

I'm not totally clueless, but things can get complicated *very* fast.
Yes, we do have to worry about dependencies and rebuilding because of
the prevalence of GPL-licensed code. All FD packages nowadays contain
both binaries and source code.


_______________________________________________
Freedos-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-devel

Reply via email to