A swiki editor

Daniel Vainsencher danielv at netvision.net.il
Tue Jul 22 17:25:23 UTC 2003


Lex Spoon <lex at cc.gatech.edu> wrote:
> Daniel Vainsencher <danielv at netvision.net.il> wrote:
> > However, these things should not be local to the image, and shouldn't be
> > specific to code. That is, everything I write - a mail in Celeste, notes
> > in a workspace, code - should be saved to an external repository all the
> > time, and that repository should not be dependent on the specific image,
> > and browsing through it should not be as painful as looking at the
> > changes file is. 
> > 
> 
> That's quite a difference of philosophy we have. 
You're interpreting me a bit too strongly here...

 I really like using
> objects, and thus I try to stay within an object engine as much as
> possible.  
Me too - and if I wished to assume a GemStone DB, I would be completely
at ease doing it, too. Unfortunately, as I dont (I want to be able to
share the code with many other people, some of which won't want this
depedency), and I do care about recoverability, I want to be sure that
everything I do is logged to the the most stable storage I can assume -
aka flat text files. Same as changes files, Celeste storage, so forth
work, except access to the data should be easier than with code in cases
of failure or image switch.

Of course the everyday interaction must be with objects. Duh... ;-)

> It seems like a losing game to try and make external files
> interact nicely in an OO context.
Until we have GS-like ultra stable object repositories, memory protected
from my flights of fancy, it doesn't seem like that bad an idea to me.

> Just use objects, IMHO.  If you really want things not to be local to
> the image, then make a way to externalize and share the objects.  This
> should be easier than the other way around, where you try to make files
> act like objects.
When I say sharing, I mean between people. The reason I want a
repository for personal information is less for sharing (it IS personal,
after all), and more for quick recovery and not worrying about "that was
in that image".

> I don't even tend to want to share objects that much, though.  I have
> one image with almost everything in it.  When I upgrade (maybe once or
> twice a year), I export any important projects and changesets and load
> them into the new image.
> 
> The image *is* a repository.
I want to update my image all the time (and generally play with the
latest toys everybody else shares), and I don't want to lose time with
clumsy recoveries.

> > [Using the class as the model and Browser as UI for personal
> > information]
> > Yes, I like that idea better, I've thought about it in the context of
> > papers - that it would be nice to write papers in a browser, overriding
> > classCompiler so that methods are treated appropriately (for example,
> > converted to LaTeX and passed to the native tools for rendering when I
> > ask for it).
> Yes, though you can also just use comments and not need to replace the
> compiler.

It has some minimal-hack-to-produce-a-result appeal. Even better - I can
use strings instead of comments, and then cmd-E gives me instant (though
imprecise) "search across docs" power. The recovery still isnt pleasent,
but it can be improved by changes to message lists. This direction does
reuse all the work put into making the code editing cycle quick and
safe, that does have it's advantages, no doubt.

Food for thought.

Daniel



More information about the Squeak-dev mailing list