Re: Elisp addiction not as bad in light of Linux forkoholism

[Marcin Borkowski]

On 2014-11-30, at 18:35, Emanuel Berg wrote:

> Marcin Borkowski <mbork@wmi.amu.edu.pl> writes:
>
>>> It is simple: Don't fork - program.
>>
>> Wow, it's rant time!
>>
>> So you're in for a treat: I have a spare minute, and
>> let me share a story with you. (I guess I read it in
>> some interview, I don't remember now.)
>
> ...but for a lot of stuff you seem to have an
> excellent memory.

Thanks.  BTW, I was imprecise: the DEK story I read was just the first
two sentences, the rest was my stream of consciousness;-).

>> When DEK coded TeX (and published the cource code),
>> he thought that many people will actually customize
>> TeX (= the engine) to their needs. It turned out
>> that (apparently) the macro programming was more
>> powerful than he expected: almost nobody did that,
>> people did wonderful things at the macro level,
>> without ever touching the source code (apart from
>> increasing the memory constraints, which required
>> recompliation back then). This includes not only
>> LaTeX and its styles (later: classes and packages),
>> but also a BASIC and Lisp interpreters, a few
>> numerical engines, a regex engine (recently), an XML
>> parser and much more. (This is, in fact, an
>> oversimplification; some of these things require
>> e-TeX, which is a relatively small extension to the
>> engine.)
>>
>> The real hacking on the underlying engine did
>> happen, of course, but not that often. Most notably,
>> we have e-TeX, pdfTeX (which is great), pdfeTeX
>> (which combines both of them); then we have XeTeX
>> (originally only on Mac, now also Win and Linux),
>> Omega and Aleph, and - most recently - LuaTeX (which
>> is the most serious modification, and a very well
>> designed one AFAIK). (There were admittedly smaller
>> extensions, like encTeX, but they could be
>> technically just patches, not "forks".) Not really
>> that many "forks", for a program more than 30 years
>> old. Especially that eTeX and pdf(e)TeX are not
>> considered forks now, rather legal successors
>> (hardly anyone uses the original tex engine
>> nowadays), and LuaTeX gains more and more traction;
>> some (me included) hope that it will mostly replace
>> the more conservative versions some day. (LuaTeX is
>> AFAIK the only one which took the idea of giving TeX
>> really new things seriously.)
>>
>> (Well, there was also NTS, but it was really a
>> clone, not a fork, and it is almost "evaporated" in
>> Orwellian sense (even the sources are nowhere on the
>> 'net!) - go figure.)
>>
>> I guess it is a bit similar as in the Emacs world.
>> If you make a program flexible enough, people won't
>> fork it too much - they just won't need it.
>
> That's absolutely right. But the question is: do
> people really need to fork Debian just to use init
> instead of systemd? init, of course, was used in
> Debian until very recently (I first saw systemd on a
> 3.17.1 kernel). If indeed impossible, Debian is in
> part to blame.

No idea.  I'm just a humble mathematician, I know next to nothing about
OS inner workings.

> And there is no doubt that the Emacs C/Lisp
> architecture really makes extension smooth (I can't
> think of any better way) - just type the code and
> evaluate, using the same software, with immediate
> effectuation - no need even to restart the program,
> let alone recompile the whole thing.

Exactly.  Not the same (but similar) with (La)TeX; we TeX users are so
accustomed to running TeX aagain and again on a file that we may
subconsciously treat it as an interactive process;-).

> That said, I think it *is* very possible to get init
> to work on the most recent Debian releases as well. A
> distro is by definition just a way of putting many,
> many things together. Forking just to replace one of
> those puzzle pieces with another is like blowing up
> the terrorist camp to free the hostages.

Fair enough.

>> (The existing forks solved some /real problems/:
>> 8-bit-ness with Omega, complicated dvi->ps->pdf
>> route with pdfTeX, limited registers and other
>> constraints with eTeX, impossibility of RtL
>> typesetting with Omega and XeTeX, lack of access to
>> system fonts with XeTeX, problems with advanced
>> programming and other things with LuaTeX.)
>
> There is a lot of LaTeX in you post. Consider posting
> it on comp.text.tex or on you home page, if you have
> one.

Well, I guess it is more or less common knowledge among experienced TeX
hackers.  And newbie TeX users don't care.  (You know, these kids of
today...;-).)

Shameless plug: http://mbork.pl .  Quite some (La)TeX stuff, and mainly
Emacs stuff lately (postings on average once a week, usually Saturdays).
Some of that in Polish, but I switched to almost exclusively posting in
English.  (Sadly, this switch temporarily eliminated one of my favourite
topics where I have no enough knowledge of English vocabulary; will work
on that.)

> Yes, I use xelatex on Linux to compile LaTeX.
> Previously I used pdflatex but I changed because of
> some Unicode issues.

Exactly.  There is inputenc, but it is prosthesis, not a "real"
solution.  (It works by messing with active characters.  Those of you
who know a bit about low-level TeX know how dangerous it is, especially
when you write to external files, like, you know, .toc.)  That's why
XeTeX (or LuaTeX) is a good idea nowadays.  (That said, I mostly use
pdfetex myself.  LuaTeX is too slow, and I don't really need much more
than pdfetex + LaTeX + inputenc, usually.)

BTW, Unicode has its own share of problems.  This:
http://www.fileformat.info/info/unicode/char/fe18/index.htm is not the
most serious, but probably the funniest.  A more serious one is mixing
German closing quotation mark with English opening one (IIRC): they
indeed do look identical, but should have different bounding boxes.  And
this is the cutest I know of: look closely at
http://www.fileformat.info/info/unicode/block/mathematical_alphanumeric_symbols/list.htm
and try to guess what happened to "Mathematical italic small h
(U+1D455)".  (Solution: it is at U+210E.  How sweet.)

> I think it is natural with a couple of parallel
> versions/dialects/implementations for huge software
> systems. That has always been the case. But Linux
> distributions? I can't give you an exact figure, but
> it is several hundreds.

Yes, this is craziness.  Maybe it's because they are too monolithic?
There was once PLD Linux which claimed to be more flexible, but it was
extremely difficult to use.  (Though its package manager was pure
genius, rpm-based but with a /fantastic/ interactive shell.)

> And it makes even less sense as Linux is a basically
> non-interactive kernel. I don't see why you can't just
> use it to run whatever software you like?

In theory, yes.  In practice, this is more difficult.  I mean, Emacs on
Ubuntu should be as simple as sudo apt-get install emacs; but then, in
most points in time, this got you an ancient version.  The same with
TeX, and other things.  So step by step you install more and more things
manually, and then why not get Gentoo?

>> Just my 2 cents.
>>
>> (And re: Debian vs Ubuntu, I never used Debian, but
>> Ubuntu is a huge disappointment: it has been less
>> and less usable recently (especially compared to,
>> say, five or seven years ago), and it will be kicked
>> out of my machine when I have a few spare days to do
>> a reinstall.)
>
> Typically, the coolest and most experienced people use
> Debian :) I never used Ubuntu but most people I meet
> who are Linux users use it. So it can't be that bad.
> It is oriented to desktop people, though now Canonical
> wants the mobile market as well. Ubuntu is in a way
> Debian + Apple: the system is basically Debian but on
> the top they have been very active with polishing, and
> to the left and right, "lifestyle marketing" has not
> been neglected.

Debian might be a good idea, but I feel more and more inclined towards
Fedora (or Arch, on days I'm feeling more bold).

> Also remember that the Debian fork Ubuntu has been
> forked many times for similarly questionable reasons:
> Kubuntu (to have it in KDE instead of GNOME), Xubuntu
> (ditto Xfce), and so on. (Sometimes I think the WM
> developers do that just to market their software.)

Yes, I even used some of them for some time.  My goal is to set up a
decent tiling WM (preferably StumpWM, maybe Awesome), so I'm not really
interested in this KDE/Gnome/Unity/whatever dispute.  (The main goal is
to have Emacs occupy the whole screen, without these stupid decorations,
and get rid of the mouse/touchpad.  I hardly ever use anything but
Emacs, a terminal, Evince and Firefox or Chrome anyway.)

Best,

-- 
Marcin Borkowski
http://octd.wmi.amu.edu.pl/en/Marcin_Borkowski
Faculty of Mathematics and Computer Science
Adam Mickiewicz University