Re: [O] We're doing it wrong. [WAS]: Zip utility on Windows for ODT exporter

[Loyall, David]

> From: address@hidden On Behalf Of Achim Gratz
> Loyall, David writes:
> > And that's why civilized programs don't depend on external executables
> > from $PATH.
> 
> Then practically all programs are uncivilized, especially when considering 
> that
> dynamic libraries are just another form of external executables.

Yes.  But would you grant me that this is done in a more orderly fashion?

> > Now, I'd imagine that some people have argued in the past that org
> > shouldn't depend on external executables.  Clearly those arguments
> > have failed.
> 
> I'm sure that if you could point to an Emacs package that allows to work with
> archives without depending on external executables it would be used
> instead, but I'm not aware of any such package: ox-odt uses arc-mode for
> unzipping (which in turn uses call-proc for actually doing it) and then 
> call-proc
> itself to do the zipping.

I realized shortly after my post that calling external executables is the norm, 
not the exception.

Also, I must apologize, my general tone in that message was terrible.  I'm 
experimenting with quitting smoking.  Suggestion: never start.
 
> > But, let's take a fresh look.  How about this rule of thumb: don't
> > depend on external executables **from $PATH**.
> >
> > Can we agree on that?
> 
> No, because I can't really see the point, especially since Emacs doesn't use
> just $PATH for call-proc, but a user option exec-path (whose default value is
> a copy of $PATH, but even a cursory look on $PATH on a Windows system
> should convince you that you really should change this).
> 
> > How about: don't depend on external executables from $PATH, but allow
> > the user to override via config.
> 
> How about: if you want that level of control, customize exec-path (and
> perhaps exec-suffixes)?
> 
> > This is important on the 'reproducible research' front.
> 
> Are we still talking about Windows?

No.  Well, kinda.

> You'd need an audited system if you
> want to take it that far, I'm not sure anybody has tried to do this on Windows
> and is still outside the asylum.  The only practical way seems to deliver the
> reproducible research as a VM (yes, that has other problems).

Yeah, I've thought about that a little bit.

I heard somebody say the other day that according to some survey, x percent of 
people don't know the difference between a search engine and a browser.  Would 
they know the difference between an application and a VM that auto-starts an 
application?  ...If you just change the title bar of Virtualbox to say "Emacs" 
instead...

I wrote ~2200 characters on this subject, just now, but then I stashed it away 
rather than present it here before asking: has this been proposed before? What 
was the outcome?

> Regards,
> Achim.
 
Cheers,
--Dave