[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kfm-devel
Subject:    URLs
From:       Stephan Kulow <coolo () itm ! mu-luebeck ! de>
Date:       1997-07-12 20:41:06
[Download RAW message or body]

Hi guys!

While fiddeling with files with '#', I found a little problem
with the definition of an URL:

I've created a directory called "http:" with a subdirectory
"www.test.de" and a file with in called "test#section?%23" and
I made some tests, what application handles this right.
- vi has no problems with it :-)
- if I "open file" within netscape, it says "can't open test",
 that means, it has the same problems like kedit (below)
- if I open file:/home/coolo/http:/www.test.de/, it opens the file
 as: test%23section%3F%2523, that means he escapes the unsave chars.
 (this is RFC1783 conform!)
- if I run `kedit http://...` or even, if I run `kedit ./http://...`,
 kedit thinks, it's a URL, unless I call it with /home/coolo/http:...,
 but then kedit has problems with the # in the name.
 Only to clearify: this behaviour is correct related to RFC 1783, 
 since the # is NOT part of the URL, but the problem is, I haven't
 entered a URL, but a filename! (kedit is of course a place holder 
 for almost every app, that uses kurl and libkfm)
- kfm shows this directories and the file correct (with my patch),
 but I can't open or drag this files with the URL meaning.

Torben, you said, you would change the way, filenames are handled,
to make sure, they're not misinterpreted, but I think, the problem
is a little deeper. Because then you have the same problem with 
calling kedit as I had. I would suggest to encode the filenames like
described in the RFC, but they have to be displayed undecoded. I don't
understand the KFM code that much, as I would volunteer here ;-), but
I would suggest to add a flag to kurl, that allows automatic encoding
and a method within kurl, that allows automatic decoding URLs.
AFAIK kedit looks at his argument and if kedit founds ":/" within
the argument, he thinks it's an URL and let kfm open this file.
I can't stop this only with a starting /. Of course my case is almost
unnatural, nobody will have a director called http:, but he could 
have one, that has a ':' as the last character of a directory name.

I think, a test, if the argument starts with a supported protocoll,
is better, but this shouldn't coded by every app, it should be better
done within kurl (or a new class, since not every app supports every
protocoll). If the app thinks, it is a URL, he should call kurl with-
out encoding, but if this isn't a URL, it is a must to encode the
argument. And if the app (or kfm) wants to handle the file ment by
the URL, it must be decoded. I can program the needed code from kurl,
but this has no sense without this is used first within kfm and second
in the apps. 

Finally, I think, the '#' is a bad (known) bug within the whole net
system of the KDE, so we should fix it ASAP. But I understand, if this
isn't the top of Torben's TODO list, since it maybe some seldom cases,
in which this makes trouble, but AFAIK emacs stores temporary files
as #file# and you can't open this one with dragging it from kfm to
kedit (for example).

Greets, Stephan

[prev in list] [next in list] [prev in thread] [next in thread]