Austin Donnelly <[EMAIL PROTECTED]> writes:

> > my whole point was that we should try to come up with a reasonable
> > interchange format for multi-layered images instead of using XCF
> > which isn't really well-suited for this task. Introducing XCF support
> > into various other apps will make that even more difficult. Perhaps
> > I'm thinking too idealistic here...
> What's so bad about XCF anyway?  It's got a version identifier, and
> it's a tagged format so old loaders can skip sections they don't
> understand.  The only problem is the fixed tile size, yes?

not really. OK, the file format does allow additions by adding new
property tags. These tags are integers which will lead to problems
as soon as XCF is not only used and extended in The GIMP. Imagine
another application wanted to use XCF since it's such a great and 
widely understood format for multi-layered images, but needs to add 
some information which is not yet there. If they'd extend the property 
enum, they'd risk that someone else used the same tag for another 
addition. Since current XCF lacks a lot of features, extensions are
very probable. With the current scheme, I imagine the whole thing 
will get ugly very soon.

> Representing the same information in XML or whatever the sexy standard
> is this month doesn't buy us anything over what we have now.

Hmm, a format with properly namespaced tags would at least avoid the
problems I've outlined above. It doesn't need to be XML, though XML
would definitely not be a bad choice to define the structure of an
image and all the metadata people might want attach to it.

> Especially since the code used in loading and saving XCF is by now
> fairly mature and (hopefully!) bug-free.  Re-coding it is only going
> to introduce new bugs.  See the recent article linked off Slashdot
> about why throwing your code away and starting again is a bad idea,
> using Netscape and MS Word as examples of large projects which tried
> to re-code from scratch and failed: Netscape went bust, and MS quietly
> canned the Word re-write project.
> We should learn from the mistakes of others :)

I personally don't think the Netscape people made a mistake here and
I do believe that carefully rewriting an app piece by piece is the 
best thing that can be done for most apps.

Salut, Sven
Gimp-developer mailing list

Reply via email to