Trash mechanism

[David Faure]

On Wednesday 25 August 2004 14:10, C. Gatzemeier wrote:
> Am Wednesday 25 August 2004 13:05 schrieb Mike Hearn:
> 
> > LD_PRELOADed libraries can be useful but we need better infrastructure
> > in order to deploy them as part of a desktop system (in particular, some
> > way to mark binaries as having particular libraries preloaded).
> 
> Yeah, I had simmilar doubts. But after reading libtrash readme I found it 
> quite reasonable. Libtrash already takes several measures. When preloaded it 
> checks for environment variables that can control its behaviour for example.

I don't see libtrash as a reasonable solution for the needs of desktop-level trash
functionality, for two reasons:
* Bad design
Changing the meaning of low-level glibc calls is not only fragile, it's also unflexible.
How would one be able to ensure that manually-deleted files from /tmp that could
be important to the user get to the trash, but temp files from compilation don't?
The proper solution is to let the low-level calls do what they always did, and let
the user-level applications (GUI apps and fileutils tools) use the trash functionality,
while other programs deleting files (like gcc for the case of object files) shouldn't.
* Lack of metadata information
We need a proper trash abstraction with metadata associated to the deleted files 
(including date of deletion, original mtime, original location),
of which libtrash only provides the third one.

As always: let's try to standardize on the behavior, not necessarily on the code.
If we come up with a reasonable standard for this, libtrash can obviously use
the same on-disk representation as the GUI filemanagers, for people who really 
want their command-line tools and other apps to trash files automatically.

What I want to know is what Nautilus and other filemanagers with non-trivial trash-can
functionality currently do, to come up with a proposal that doesn't diverge too 
much where not necessary. (Unfortunately the Nautilus developers who took part in
the 2001 discussion were eazel developers and now work at apple...)

-- 
David Faure, faure at kde.org, sponsored by Trolltech to work on KDE,
Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).