Re: What's your favourite *under_publicized* editing feature ofEmacs?

[Eric Abrahamsen]

On Mon, Feb 28 2011, Perry Smith wrote:

> On Feb 27, 2011, at 9:28 AM, Cthun wrote:
>
>> On 27/02/2011 3:08 AM, David Kastrup wrote:
>>> Cthun <cthun_117@qmail.net.au> writes:
>>>> Oh, really? I for one cannot recall ever seeing a version 1.5 of a
>>>> novel or a version 2.0 of a magazine article.
>>> 
>>> Well, I've been responsible for the typesetting software for "Die
>>> Kritische Gesamtausgabe der Werke von Ernst Troeltsch"
>>> [anecdote trimmed]
>>> 
>>> So definitely there were various versions of the same article published.
>> 
>> But that's not the same thing as software versioning, or anywhere close.
>> 
>>> The gestation of both articles and novels is rarely linear.
>> 
>> True enough. But it is also not going to fit especially well to what
>> systems designed for software revision control do. There is a single
>> long piece of text rather than lots of interacting software modules,
>> for one thing; there are no builds or library dependencies or bug
>> reports or feature requests. There's also a point where it's
>> actually *finished*, while software is never finished and has many
>> successive versions released, each fixing the bugs in the previous
>> and adding new features.

[...]

>> Then woe betide you if you ever work on, say, a novel rather than a
>> computer program.
>
>
> The phrase you pulled out referred to "version control". You are now
> talking about source code control systems...
>
> Also, SCMs that I know like git, svn, rcs, bzr, etc do not have any
> concept of build, dependencies, bug reports, or feature requests. Its
> one of my frustrations. If you know if a single system that has all
> those, I'd love to know about them. The only fully integrated example
> I have is IBM's CMVC and IBM dumped it because no one understood it.
>
> Aside from that, I'm not getting your point at all. An article can
> easily be broken into sections, a book into chapters which are then
> subdivided into sections. Those sections will have dependencies. I
> doubt if the author will add those in but the concept still applies.
> And as far as history, there are countless examples where a single
> journalistic piece has a very long life to it. Haven't ever listened
> to where historians go back and review the author's original notes?
> God... there are entire books about the American Constitution trying
> to reconstruct the various versions and the intentions behind each of
> them.

To add my cent-and-a-half… I use emacs (and git) for novel
translation—functionally the same as novel writing. While I'm far
happier with this setup than with any other (in moving from a Mac to
Linux, my only regret is the loss of Tinderbox), I can certainly see
cthun's point. When you are writing long-form text, the unit is the
paragraph. When writing code, the unit is the line. Writing prose, the
addition of one word can transform a whole paragraph (using fill-mode).
Writing code, the addition of one "word" generally only changes a line.

Version control systems thus become that much less useful. Not useless,
just less useful. Git is still great for doing a commit with translation
and research notes in, and then another with them out, or even
branching—that sort of thing. But you have to think harder about how
conceptual alterations should be recorded as version changes.

E