Re: sqlite3

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> Application state is indeed a different issue, that should be decided
> separately on a case by case basis.  For example, should
> .emacs.desktop remain in its current text form?  I personally think it
> should, because it's still readable enough and performant enough, but
> there could be other opinions.

Yeah, I think .emacs.desktop is probably fine as it is.

Stefan M raised the issue about savehist -- the problem there is that is
you have two Emacsen running at the same time, the one you close last
will overwrite the data from the other instances.  Quintan's log
approach might work for that, but I'm wondering what the semantics
should be.

That is, ideally I think if you do A in Emacs 1, B in Emacs 2, and C
in the first Emacs again, the savehist should have

A
B
C

and not

A
C
B

(if you closed the first Emacs first, and then the second).

Possibly.

And perhaps...  after doing A in Emacs 1, you might be able to see that
immediately in Emacs 2?  But that could also lead to all kinds of
usability problems.

So it's no clear what the interfaces here might be for these more
complex application states, and there may not be a "one solution fits
all" approach for these.

> Large databases are yet another different beast.  For example,
> consider the mairix-related features in Emacs: we current use an
> external program for that, where using a database internally from
> Emacs could make at least some sense: for starters, we wouldn't be
> limited by search capabilities of mairix the program, but could
> instead unleash the full power of Emacs searching capabilities.

Yup.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no