Hi,

The difference you missed (actually I missed this as well as I don't use this constructor:) is that public HttpRequest(string filename, string url, string queryString) of MS.NET seems to initialize QueryString to a NameValueCollection using Encoding.Default as a hard coded encoding (when I changed default code page in Control Panel the encoding used changed so this is Encoding.Default for sure, and no configuration is used). When HttpRequest is created from a HttpWorkerRequest query string is processed only when QueryString property is accessed for the first time.

As query string is processed only once, changing HttpRequest.ContentEncoding has effect only if you change (either programatically, in configuration or using HTTP headers) it before accessing QueryString.

So query string parsing should be moved to a separate method that takes encoding as a parameter. And call with Encoding.Default in the public constructor and with HttpRequest.ContentEncoding in QueryString (and of course only when it was not yet parsed).

Kornél

----- Original Message ----- From: "Juraj Skripsky" <[EMAIL PROTECTED]>
To: "Kornél Pál" <[EMAIL PROTECTED]>
Cc: "Miguel de Icaza" <[EMAIL PROTECTED]>; <mono-devel-list@lists.ximian.com>
Sent: Monday, May 08, 2006 1:41 PM
Subject: Re: [Mono-dev] Patch for HttpRequest.cs


Hello,

I was talking about the encoding used during the URL decoding only. My
patch fixes that. Running the attached test program demonstrates the
need to call HttpUtility.UrlDecode with Latin1 encoding to match MS
behaviour. No matter what encoding is set in
HttpRequest.ContentEncoding, MS.NET always URL decodes "%e4" to "ä", so
it must _always_ be calling

HttpUtility.UrlDecode("%e4", Encoding.GetEncoding("latin1"))

Or am I missing something? Any feedback appreciated!

- Juraj


On Mon, 2006-05-08 at 12:57 +0200, Kornél Pál wrote:
Hi,

You are wrong. HttpRequest.QueryString does the following on MS.NET:

The only encoding it uses is HttpRequest.ContentEncoding. It tries to obtain HttpWorkerRequest.GetQueryStringRawBytes(). If it fails then falls back to
HttpWorkerRequest.GetQueryString(). When it was able to obtain the byte
array it will decode it using HttpRequest.ContentEncoding.GetString(). As
such query string is decoded correctly. When no byte array is available in HttpWorkerRequest or the query string was set either in constructor or using
HttpContext.RewritePath for example the string is assumed to be decoded
correctly so no decoding is done.

Now we have a string that still may be URL encoded. MS.NET probably calls
HttpUtility.UrlDecode just like we do but MS.NET passes
HttpRequest.ContentEncoding as well because query string is assumed to be
URL encoded using that encoding.

Note that obtaining query string from HttpWorkerRequest in the constructor as we currently do is a wrong implementation as HttpRequest.ContentEncoding
can be changed before HttpRequest.QueryString is first accessed.

We should do the following:
- delay query string processing until it is needed (don't obtain query
string in the constructor)
- try HttpWorkerRequest.GetQueryStringRawBytes() as well
- use HttpRequest.ContentEncoding to decode the byte array and for
HttpUtility.UrlDecode

Kornél

----- Original Message ----- From: "Juraj Skripsky" <[EMAIL PROTECTED]>
To: "Miguel de Icaza" <[EMAIL PROTECTED]>
Cc: <mono-devel-list@lists.ximian.com>
Sent: Monday, May 08, 2006 12:22 PM
Subject: Re: [Mono-dev] Patch for HttpRequest.cs


> Hello,
>
> After running more tests, I've found out that on MS.NET the decoding in
> HttpRequest.QueryString does _not_ depend on
> HttpRequest.ContentEncoding. In fact, MS seems to be always using > Latin1
> here. All other standard encodings fail.
>
> A revised patch is attached, including a NUnit test case. If no one
> objects, I'll commit.
>
> - Juraj
>
>
> On Sat, 2006-05-06 at 13:47 -0400, Miguel de Icaza wrote:
>> Hello Juraj,
>>
>> > The attached patch makes sure that the get-parameters in QueryString
>> > are
>> > url-decoded using the proper encoding (when creating the
>> > NameValueCollection).
>> >
>> > May I commit?
>>
>> Could you provide NUnit tests for this case?
>>
>> Miguel
>>
>


--------------------------------------------------------------------------------


> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list@lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>



_______________________________________________
Mono-devel-list mailing list
Mono-devel-list@lists.ximian.com
http://lists.ximian.com/mailman/listinfo/mono-devel-list

Reply via email to