Charset & encoding through CIFS / Webdav / FTP...

[jgoffinet]

Hi,

I've made some tests through several protocols. Here is the config I used:

- server: linux (debian) / mysql / tomcat bundle (1.0.0 release) - no special config

- client: WinXP Pro SP2 with :
   * Windows Explorer to access CIFS share
   * Windows Explorer with network places to access Webdav
   * Windows Explorer with network places to access FTP
   * DavExplorer to access webdav
   * FileZilla to access FTP

Here are the results of my investigations:

1. I had no success in connecting to FTP with FileZilla. Here is the log on the client side (no log on the server side but a java.net.ConnectException: Connection timed out):

– beginning of log
PORT 192,168,1,113,6,173
Réponse :   200 Port OK
Commande :   LIST
Réponse :   150 File status okay, about to open data connection
Erreur :   Inactivité détectée !
Erreur :   N'a pas pu récupérer la liste du répertoire
– end of log

(the 2 last lines mean: "inactivity detected" and "couldn't retrieve directory list").

2. Everything worked fine otherwise… until I tried to create spaces with accents in their names. I created the followings:
- "cifs-é" with Win Explorer
- "webdav-é" with network place
- "ftp-é" with network place
- "davexplo-é" with dav explorer
- "web-é" with web client
and looked at the result in the different clients.

3. CIFS, Web client, DavExplorer: for spaces created with CIFS, web client and DavExplorer, there seems to be no problem: space names are shown correctly, and it is possible to browse into them.

4. FTP (through network place): the folder "ftp-é" is shown as "ftp-?" in every client, even the one which was used to create it. It is impossible to browse into it through Network place. It is browsable only via web client and dav explorer (even if it is displayed as ftp-?). cifs-é, web-é, etc. are also shown as cifs-?, web-?, etc and are not browsable through FTP network place.

5. Webdav (through network place): the folder "webdav-é" is shown as webdav-é in every client (network place, cifs share, web client, dav explorer),but it is possible to browse into it. Other folders are shown correctly (cifs-é, web-é, etc.) but trying to access them leads to an error (folder unavailable).

Did anyone else encounter such troubles with accentuated files ?

Jean

[gary_spencer]

Hi Jean,

I'll have to do some tests with accents in names to see what is going on here.

CIFS should be ok as it supports Unicode file/folder names. FTP doesn't currently support Unicode, I'll need to look at the spec for how that is supported. WebDAV looks like it may be a conversion problem.

Cheers

Gary

[gary_spencer]

Hi Jean,

Just did some quick tests here using the latest v1.1 code.

WebDAV via Network Places seems ok at creating and viewing folders with accents in the name. Folders were created with CIFS and web-client.

CIFS and web-client seem ok at creating/viewing folders. FTP via Network Places was also able to view the folders created by CIFS, WebDAV and web-client which I was suprised at as I expected it to fail.

Could you try your tests again using the latest v1.1 code.

Cheers

Gary

[jgoffinet]

Thank you Gary for your response.

> Could you try your tests again using the latest v1.1 code.

Do you mean the latest stable 1.1 version or the latest nightly build ?

Jean

[gary_spencer]

Hi Jean,

I would try the stable v1.1.

Cheers

Gary

[jgoffinet]

With 1.1 release (linux tomcat bundle), I made this very simple test:
- open a webdav network place pointing to the repository
- add a folder named "webdav-é" via Webdav
First the folder name is correctly displayed, but if I refresh the screen I get "webdav-é" instead of "webdav-é" in the folders tree. Note that I can still access this directory.

Now another test:
- create a space named "web-é" through the web client
- refresh the folders tree in the Webdav Network Place: the "web-é" directory is correctly displayed
- click on that directory
An error message pops up, saying that the folder is unavailable, and the content of the folder isn't displayed…

I don't know if this could be linked to my Windows configuration, but I don't remember having done anything special… I use a french version of Win XP Pro SP2. Is there a way to configure which encoding the Network Places must use ?

Thanks,

Jean

[gary_spencer]

Hi,

I've managed to reproduce this problem now, not sure why I wasn't seeing t before. I'm looking into it at the moment.

Cheers

Gary

[unknown-user]

My previous issues about 1.2.1 and 1.3dev versions of alfresco seems related to those charset problems.

I experience the exact same behaviour of broken accents using webdav.

I agree with the fact that it must be a conversion problem : using WebClient, everything is fine, though my debian and mysql are using UTF8 and the client  (windows XP) is using latin9 aka iso8859-15.

Crashes, and slow operation were surely related to this (windows explorer is almost hanging when meets an accent, which sadly happens every 5 minutes with  french repository 😃 )

Where could we specify client charset? (for example webdav client or cifs client) Isnt the header supposed to announce content encoding? Windows XP bug? How could i analyse this up?

Thank you!

[lgr]

Hi,

Testing with v1.4, it seems that accents are not understood in FTP.
Webdav and cifs work well.
(using postgresql, i presume no utf8 configuration because cifs and webdav both work well)

Some Jira issues deals with accents and ftp, but are resolved as of october, 26, the day after 1.4 release.

Does it means that ftp will be working well with UTF8 in the next version ?

Laurent.

[kevinr]

The UTF-8 support in FTP fixes problems with accents/umlauts etc. in 1.4 community - but it depends on the FTP client enabling it.

Thanks,

Kevin