Redundant (?) element of an URL

[jsbien]

What #Linde is doing at the end of http://poliqarp.wbl.klf.uw.edu.pl/extra/linde/index.djvu?djvuopts&page=2444&zoom=width&showposition=0.533,0.848&highlight=477,2745,137,54,ffff00#Linde?

For this URL djview4 doesn't use the showposition parameter. Perhaps this is somehow related?

[Michał Rudolf]

I am just using what is provided by the server. As we cannot be sure how the server handles # in URL, I think it is better to keep it as it is (see the problems with redirection).

[jsbien]

This is not an URL from the server, at least not directly, but created by Copy URL.

[Michał Rudolf]

The # part comes from the server.

[jsbien]

Actually such URLs are never handled by a server, but by djview4. Nevertheless investigating the origin of the redundant part is an issue in marasca, not djview4poliqarp.

[Michał Rudolf]

First the URL is passed from djview4 to the server to retrieve DjVu file. As we allow external servers, we cannot be sure they don't use # part of URL to retrieve the file.

We already had an example of server using djview4 parameters: appending highlight color to highlight= parameter of URL broke redirection on the wbl server.

[jsbien]

The problem requires reinvestigation in some future as it is quite nasty: the highlights for added entries don't work. It does not appear systematically, I will try to find a way to reproduce it.

[jsbien]

I've added some entries to an index and the bug hits again, making the new entries unusable without some hand-editing:

Wanda;http://djvu.szukajwslownikach.uw.edu.pl/linde-t/06/index.djvu?djvuopts&page=216&zoom=width&showposition=0.541,0.788#LindeNowy&highlight=2443,1292,269,111;::018-2-32-0-0 __ _ ___::;※ a:imię
Wenda;http://djvu.szukajwslownikach.uw.edu.pl/linde-t/06/index.djvu?djvuopts&page=216&zoom=width&showposition=0.541,0.788#LindeNowy&highlight=2802,1294,321,98;!::018-2-33-0-0-s_ _ ___::;※ a:imię

[Joachim Aleszkiewicz]

Actually, I've checked using Wireshark what requests are made by djview and djview-poliqarp (2d2215e) when requesting

http://djvu.szukajwslownikach.uw.edu.pl/linde-t/06/index.djvu?arg=a&djvuopts&page=216&arg2=val2&zoom=width&showposition=0.541,0.788#LindeNowy&highlight=2443,1292,269,111

(address as jsbien's, but with arg=a added before djvuopts and arg2=val2 after)

Surprisingly the request was GET /linde-t/06/index.djvu?arg=a.

As is, neither djview nor djview-poliqarp inform server of anything after djvuopts&

Regardless of whether this is expected behavior, anything after # (that is, 'fragment' section of URI) can be safely removed.

According to RFC3986 (URI: Generic Syntax), section 3.5.:

"the fragment identifier is separated from the rest of the URI prior to a dereference, and thus the identifying information within the fragment itself is dereferenced solely by the user agent, regardless of the URI scheme"

If any server uses part after '#' to change returned resource, it is deeply broken.

[Michał Rudolf]

I don't have enough time now to reverse-engineer the server source code.

Is it possible to see a page with a link containing hash symbol?

[Joachim Aleszkiewicz]

Yes, but any options after # are discarded.

So the problem starts when adding new djvuopts, since they are added at the end, possibly after #LindeNowy which is apparently part of the search result URL.

[jsbien]

Please not the problem is specific to djview4poliqarp, it doesn't occur in djview4, so I would not be sure it is the problem of the server. Moreover the problem started to appear when the server itself was already stable. Anyway, a brute force approach consisting in removing the offending part before storing it in the index or clipboard seems sufficient.

[Michał Rudolf]

How can I obtain URL with # in djview4poliqarp?

[jsbien]

E.g. adding an entry to a dictionary.

However we can put the issue on hold, I have some new ideas which I will check in the coming days.

[Michał Rudolf]

I meant: how can I see the # link served by server or djview4poliqarp. I understand I can add such link manually, but where does it come from?

[jsbien]

Adding an entry doesn't mean adding the link manually, the link for the entry is created by djview4poliqarp.

[jsbien]

As I said, I want to check some ideas.