Shibboleth, Tomcat, AJP and UTF-8 conversion

[Mark Juszczec]

On Thu, Oct 20, 2016 at 10:43 AM, Peter Schober <peter.schober at univie.ac.at>
wrote:

> * Mark Juszczec <mark.juszczec at gmail.com> [2016-10-20 16:21]:
>
> > Yes.  The common solution is to specify one or both of the following in
> > <Connector> in server.xml:
> >
> >  URIEncoding="UTF-8"
> > useBodyEncodingForURI="true"
> >
> > Unfortunately that has no impact on my problem.
>
> Those are for data that's part of the URL, no? As such those should
> have nothing to do with request.getAttribute() calls to data provided
> outside of query parameters?
>

Based on my research I've found that these are the only ways to influence
the encoding on parameters passed in the body of a GET.


>
> Also, isolatin1 is only the "DEFAULT_CHARSET", i.e., when none is
> set. No idea when 'if (charset == null)' happens, but are you sure
> you've provided all means (HTTP, HTML) to the User Agent to make sure
> everything is UTF-8?
>
>
That's a good point.  I've gone back and looked at the
ByteChunk.getCharset()

    public Charset getCharset() {
        if (charset == null) {
            charset = DEFAULT_CHARSET;
        }
        return charset;
    }

The data returned by HTTP is fine.  Once it gets passed thru AJP, its not.

> I suspect its because my data, for whatever reason, is NOT coming to
> > me % encoded.  mod_jk is putting it into environment variables and
> > the delivering it via AJP in the body of a GET.
>
> Unrelated to your issue (which is one I think everyone shibbolizing
> Confluence ran into at some point), but last time I checked mod_jk
> didn't provide any features I absolutely needed to warrant the extra
> work of building it myself, when mod_proxy_ajp comes fully supported
> and often packaged by GNU/Linux distributions.
> It should have more features and tunables, though.
>


Can you explain "shibbolizing Confluence" ?

I don't understand.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://shibboleth.net/pipermail/users/attachments/20161020/8a9e3cd4/attachment-0001.html>