Can't really add much to Erwin's comments, other than one quick question.
What encoding is your XML file in? I've seen errors like this when using
8859 encodings, but I have not used the encoding declaration, so the
parser tries to read as UTF-8. Similarly when I have a let a 8859 char
"slip into" the UTF-8 document by accident.
Cheers,
Berin
>> Now could successfully verify signed files which had "cascaded
>> namespace attributes".
>> The reason for the earlier failure seems to be when verifying larger
>> (>40kB) signed files.
>
> There are known problems with large files, but that are only really
> memory issues because at present we need to have the entire DOM tree
> in memory. These files are usually many MB large (depending on memory).
>
>> I got another (also Apache-XMLSignature-1.0.4) based tool from a
>> colleague which gives some more detailed error messages. When
>> verifying larger files I get:
>>
>> java.io.UTFDataFormatException: invalid byte 1 -f 1-byteUTF-8 sequence
>> (0xbf)
>> at org.apache.xerces.impl.io.UTF8Reader.read(UTF8Reader.java:678).
>>
>> Is this a known issue ?
>
> It's not a known XML-Security bug, but it does look more like a Xerces
> bug than a XML-Security bug. Have you checked the Xerces bug reports?
>
>> Any idea what is wrong ?
>
> It seems like Xerces doesn't like a particular character in the XML.
>
>> When I decrease the file size by deleting elements, resign, and then
>> validate again, everything goes fine. The critical file size seems to
>> be around 40kB.
>
> My guess would be that you deleted the particular character in the last
> delete.
>
> Right now it's a little up in the air what's happening, but that would
> be my gut feeling with the information available at this time :)
>
> Erwin
