Re: Gnome 2.10 Showstoppers : the list

[Paolo Borelli <pborelli katamail com>]

Luis Villa wrote:

> On Fri, 4 Feb 2005 01:49:59 -0500, Luis Villa <luis villa gmail com> wrote:
>  
>
> > On Fri, 04 Feb 2005 07:43:43 +0100, Paolo Borelli <pborelli katamail com> wrote:
> >    
> >
> > > read http://bugzilla.gnome.org/show_bug.cgi?id=110191 and in particular
> > > the comment #20. We have (almost) the code, but we are not confortable
> > > with the testing: it will go in as soon as 2.11 is branched so to get
> > > tested for a whole developement cycle.
> > >
> > > And no, it obviously isn't a show stopper nor a regression.
> > >      
> > Yeah, absolutely not. It would be great for someone to do an audit of
> > this kind of thing and have it as a focus for 2.12, but it totally
> > isn't a showstopper.
> >    
>
> (On a non-release-team-y note, if this is so hard to do, does that
> indicate something is deeply fucked in our APIs?)
>  

Yes, API and implementation: gnome-vfs tries to use posix apis to 
transparently deal with files on remote locations. While this sounds 
really nice in theory, in practice many protocols cannot implement some 
of the posix semantics in a sane way (think of trunctate or of an atomic 
rename).
See http://live.gnome.org/GnomeVfsPlans for the beginning of a design 
discussion.

To that add the fact that many of the vfs modules didn't see much 
testing outside nautilus, thus some of the stuff which nautilus doesn't 
use is buggy. For instance, dealing with symlinks on sftp may easily 
crash the app, or trying correctly deal with permissions on methods 
different from file:// is next to impossible.


ciao
   paolo


PS: wrt to gedit, lemme note that even if gnome-vfs write didn't make 
into 2.10, we added a "Save Copy" plugin which allows to save a copy of 
the document also on remote locations (saving a copy is safer and easier 
than the real save, since it has not to deal with backups etc).