This page was inspired by Alex Gravely's BOF talk at the 2005 Gnome Summit.

Alex led a discussion on how integration between current Gnome Apps is currently pretty awful, and how it can be improved.

Alex's demo?

Topics:

• copy vs. move. How can Gnome best avoid suprising the user?
• Lack of rich formats.

• Linking between documents (Office, Abi<->Gnumeric) but avoiding the danger of broken links.

• Much to learn from Jody Goldberg's experience with Gnumeric.
• Evolution. Evolution has loads of things that users would expect to drag (contacts, calendar items, emails, etc), yet it lacks hot spots.
• Use gnome-keyring to store epiphany login and password database. Same with Mergeant and all user/password applications.

1. Discussion

Some thoughts from me...

• First off, Evolution. In my opinion, that program is currently working in a direction contrary to the usual way of GNOME apps. It is excessively large, doing a lot of stuff all in one single place. If I like Evolution's calendar but am not really a fan of its email, I still am faced with its email functionality. Distros have trouble classifying Evolution since it does so much, so they give Evolution multiple entries in the menu. Its chunks have never struck me as very tightly linked, anyway, so how about break this thing apart into some smaller apps? To make that worthwhile, there could then be some open-ended d-bus interfaces for those chunks to talk to each other (for example, the scheduler wants to send some mail informing about an event), accessible to not just Evolution but to any application on the desktop. It would be nice to see Evolution encouraging more interoperability; the current state of mail apps has the addition of another meaning the complete removal of the last.

• Something I appreciate with Apple's MacOS is that all of the default apps have a really intuitive way of finding files from other default apps. That is, the user can get his images and music without navigating through the file system. That is cool! Granted, GNOME shouldn't start imposing any particular desktop search engine, but there are ways to do this the flexible way. Distros could provide smart search folders via systems like trackerfs, so that's easy. Currently, most default programs are using very ugly directories to start in when people choose to (for example) import music. Using the XDG user dirs stuff won't cut it, because it wouldn't be cool enough. Instead, I'm thinking that settings should be in gconf for where these apps start looking (catalogued by types of data). An OS could have clever search folders elsewhere (not in the user's home directory. /media or something), visible in the Places list, and as the default paths for certain applications to look in. For example, someone using Rhythmbox looking to import music would be given a nice list of different places music can be found and be able to very easily click one of those. This can be particularly cool when the indexer is scanning the network, since then sharing data not just between programs, but between systems, becomes very seamless. Another nice gain here is that we can escape standards-hell (where everyone wants a standard place and format for every type of data, but that refuses to happen), since the only thing that needs to keep up with the different filing systems is the search engine, as opposed to every application on the desktop. Well, that last thought would also need a miracle in between since converting between formats is hardly seamless. Any day now, folks...

• Tagging. It's a big one. Evolution has some kind of tagging / categorizing system. So does Nautilus. And F-Spot. Oh, and Tomboy. I've missed quite a few. Any program which offers a "star" option does this, to some extent. With most of these, the different things we decorate with tags are actually their own files, too. Flipping through my files with Nautilus right now, they are all quite bare. However, inside those files, I have all sorts of classifications, emblems / tags / categories... Nobody can agree on a format, which is something I tend to grudgingly accept for other data, but the weird thing is these all work under the same principles. Wouldn't it be cool if we could navigate to the F-Spot library with Nautilus and see all of the F-Spot tags on those images as if they were Nautilus' emblems? (Or as if we were using F-Spot). As you may have guessed, this bullet point is further pointing out the obvious: That desktop-integrated tagging (and tagging in every default GNOME app) would be really cool, and there is indeed a system that wants to do it. (LeafTag). That would result in unmanageable heaps of tags to choose from, I hear? The hierarchical nature of F-Spot's tags are really neat, and I think would work great here. Where one program (Evolution, for example) may primarily use Office tags, another may tend more towards Utility tags. These could be filed as children of tags called Utility and Office, rather than in the root list of tags. Cool gain in functionality there, where one can tag files he wants backed up, then a generic backup tool can use the distro's search tool to find them.

-- DylanMcCall 2007-10-28 01:41:48