Re: File selector talk writeup



On Sun, 2003-02-16 at 15:52, Owen Taylor wrote:
> The idea of having the application augment the list of mime types
> for the non-MIME-system case is reasonable, though:
> 
>  - It might seem clunky on systems where extensions do define
>    file type (Windows, and apparently now MacOS)

When you add a type, you would of course have to specify that through
the extension and (possibly) the magic binary sniffing.  On those
systems that always use extensions exclusively [1], GTK could avoid
using the binary sniffing completely when getting the MIME type, for
consistency with the rest of the system.

> - People may want to use other patterns in some cases ...
>    e.g. *~ to show only backup files or something.

And I don't think the second case is common enough to make it necessary
in the API.  The *~ case doesn't even happen at all with
standard-behaving desktop apps (i.e. no backup files).

> Yeah, it is pretty much a UI decision. If we don't add filter selection
> people will doubtless ask for it, but doesn't mean that it is right.

My point exactly.  :-)

> As long applications allow you to export as a filetype, we aren't
> going to be able to hide the idea of filetypes from the user, though
> that may make sense on theoretical grounds. If I have a directory
> full of .xcf files and also .png files that I've exported from them,
> viewing only the .xcf files may make sense.

Or if you have been smart enough, the XCF files go into a different
folder. :-)

> Is "Open Read Only" really that generally useful? I brought it up
> as an example because it was mentioned before, but it's not something
> I've seen very often, and certainly not something that I've wanted
> to do very often.

This is not something I do very often either.  Actually, if it boils
down to it, I wouldn't add that widget to the API either.  :-)

For the "extensions" to the dialog, they really are a UI decision, so
every addition should be evaluated as "sane" UI-wise before hooks are
added for it.

That's why I am opposed to a generic add-wacky-custom-widget API.  It
might give us a theoretically less versatile API, but in practice, it
doesn't matter; after all, programmers worldwide have been surviving
with the "limited" APIs of Windows and MacOS all these years...

> > Besides, for a such a common feature it makes sense to have an API
> > shortcut that doesn't involve creating the widget manually
> 
> Hmm, I'd be interested to see what your API proposal for this is.
> It seems to me that if we have 4-5 things of the type:
> 
>  gtk_file_chooser_set_show_open_read_only()
>  gtk_file_chooser_get_show_open_read_only()
>  gtk_file_chooser_get_open_read_only()
>  gtk_file_chooser_set_open_read_only()
> 
> Plus corresponding properties it's going to make for a big
> messy API.

If an optional element is considered to be important for the UI, I think
that's a price worth paying.

Still, I really think we should have a very limited set of those
optional items (if at all), so that wouldn't make for a big API in the
end anyways.

[1] Which BTW is far better from what we do on Linux, where we respect
the extension on odd days and do binary sniffing on even days.  :-)

-- Ettore

Attachment: signature.asc
Description: This is a digitally signed message part



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]