Custom URI schemes & KDE

[Frans Englich]

On Friday 30 December 2005 21:50, David Faure wrote:
> On Friday 30 December 2005 17:39, Frans Englich wrote:
> > An application needs to shuffle data around and -- for internal use --
> > invents a "data" scheme. Then the application grows and the URI scheme
> > becomes part of, say, an interface between plugins. Who knows what
> > happens, perhaps KIO gains support for the actual data scheme(RFC2397)
> > and the mess is there.
> >
> > If that imaginary example seems like a strawman, perhaps the KMyMoney
> > case doesn't. The scheme name "sql" was decided for, which more or less
> > begs for that Oracle, IBM & Acme writes an RFC for the "sql" scheme,
> > which then a DB related class in Qt/KDE gains support for ... and then
> > the confusion and potentially technical problems are there.
>
> Well it's not too hard to name it "kmymoney:" instead of "sql:" to avoid
> this.

You may think so, but that didn't prevent it from happen. For example, there 
was no doc on developer.kde.org which said 'avoid inventing schemes'. 

And the advice "Yeah, just invent a scheme and use your app name" is bound to 
cause trouble. Right, "kmymoney" has a low /likely/ hood, but what about 
"kgdb"? And what is the likely hood in X years?

Could you explain to me why the tag or URN scheme exist at all? Is the W3C and 
IETF people simply wrong, since they go through all the hazzles of 
registering schemes, inventing tag & URNs? "What's wrong with private URI 
schemes?"[1]

Inventing schemes can lead to problems -- theoretically if you so like -- but 
give me a reason to not do it properly.

> > PS. I think "normal-looking protocol" is a weird term,
> > "tag:kde.org,2004:KMyApplication" is a fully valid protocol. One could
> > easily use something like that with KIO.
>
> Right, and we'll have a file named "kio_tag:kde.org,2004:KMyApplication.la"
> to implement that kioslave? Good luck with that on Windows...
> Honestly, this doesn't make sense. If you write a kioslave for something,
> then it's very likely useful to more than one application, and then naming
> it after one application [and a year!?] doesn't make sense anymore.

It was an example. We're drifting of topic; the thread is about the general 
problem of URIs, not whether some overhaul in KIO is implementable.

I don't see your point; let's return to the document. I understand that you 
would invent a URI scheme rather than follow the recommendations done by the 
W3C & IETF, but is there somekind of trouble if those who not wants to do it 
the proper way?

Would it be ok with you if people decided to use the tag scheme with the 
kde.org domain and organized/synchronized their usages in a document on 
dev.kde.org, which also is a central point with links/advices for solving 
this kind of problems?


Cheers,

		Frans

PS. Happy New Year!

1.
http://esw.w3.org/topic/UriSchemes