Non-Linux hidden files like Thumbs.db should be treated the same as .filename

Thank you for taking the time to report this bug and helping to make Ubuntu better. This bug is known upstream, the address is: http://bugzilla.gnome.org/show_bug.cgi?id=311010 ; please feel free to post your comments there too, thanks a lot.

Is there a way to read the "hidden" attribute from filesystems like FAT and NTFS that support it?

Found a thread about it: http://thread.gmane.org/gmane.comp.gnome.nautilus/12197

Seems like this needs to be added to the kernel drivers to request the hidden attribute.

And another thread on the kernel list:

http://www.spinics.net/lists/linux-fsdevel/msg17861.html

How do we mark this bug as affecting the kernel?

Here is a patch that will allow the patterns that are hidden to be changed in gconf.

Could you add your change to the upstream bug?

It's on the upstream bug; it has gone ignored for a while though. I submitted it to the nautilus mailing list as well and once again received no response at all.

Could this patch please be applied to the downstream nautilus? I don't know of any other way to get it applied upstream. Perhaps this way it could eventually work its way into upstream nautilus?

Hmm, seems the kernel devs where not so happy about the idea. If we really want this it should be done in gvfs (with filesystem specific checks) and not in nautilus, definitely not with blacklists.

So does this mean that nothing will be done at all?

Besides, I can think of cases beyond Thumbs.db, where it would be nice to have a way of hiding files. Pesky .pyc files are a good example. Nautilus already does its own hiding of files, such as files ending with ~, so why is this a big problem?

Also, sometimes folders will get copied from one filesystem to another, and so hiding Thumbs.db on a filesystem that looks like Windows, only to have it appear on an ext4 partition after copying, doesn't seem like a much better solution. Of course I'm not sure if that's what the "filesystem specific checks" would imply, but it's a potential problem.

gvfs could add custom layer to hide files using the "standard" .hidden file in each directory.

Remember there are two possibilities:

1. Reading the hidden attribute from filesystems that support it
2. Pattern-matching on commonly-hidden filenames like Thumbs.db, desktop.ini, ~*.tmp, __MACOSX, etc. - like a global .hidden file

The kernel devs refuse to consider #1, while dfalk's patch seems to implement #2.

Does Samba do something like #1 that could be implemented for local filesystems?

I don't see anything on http://bugzilla.kernel.org/ about this. The mailing list discussion says it will be ignored, but maybe we should file it anyway?

I don't think #2 is a bad idea at all, though, and could probably be extended to be part of a related idea to allow for files to be "attached" to other files, with the attachments hidden by default (book.txt~ attached to book.txt, and web_page_files attached to web_page.html, for instance). That might require glob-matching as well as filename matching, though, but it's also something done on a file-level rather than a filesystem-level.

Does the .hidden convention allow for any type of wildcard or regexp? I can't find documentation.

thank you for your work, unsetting the patch status though since that require extra upstream discussion and nautilus might not be the right component to change there either since the fileselector etc have similar issue

Solution #1 wouldn't work for hiding python .pyc files - so it might be good to go with patching nautilus (and gtk+ file opener?).

I like this bug. Good idea!

Status changed to 'Confirmed' because the bug affects multiple users.