Re: SVG 1.2 Comment: image/svg+xml;charset=""

* Chris Lilley wrote:
>Yes, when I read that I had not realized that the document had been
>carefully constructed to avoid using byte sequences illegal in either
>encoding. So the content would 'merely' be mangled, rather than
>producing a well formedness error. In the general case that won't
>happen, which is why I asked what you would do when reading from local
>disk.

This is actually not uncommon if you attempt to use e.g. ISO-8859-2 in
your XML document but the server or server-sided applications such as
PHP are configured to supply ISO-8859-1 in the charset parameter if no
other charset parameter has been specified. And your question does not
really make sense, you are essentially asking about a case where you do
not supply all necessary information to applications, which will break
no matter what. As a counter-example take

  http://lists.w3.org/Archives/Public/www-archive/2004Nov/att-0029/02573376332096d26282fa08a240a85a.svgz

This is a gzip compressed resource with a header

  HTTP/1.1 200 OK
  Connection: close
  Date: Thu, 25 Nov 2004 06:03:35 GMT
  Content-Length: 913
  Content-Type: image/svg+xml;charset=us-ascii
  Etag: "1fopdkh:101gg0pl0"
  Last-Modified: Mon, 22 Nov 2004 15:57:37 GMT
  Server: Jigsaw/2.2.5

If you ignore the charset parameter for a second (it has been added by
W3C's broken list archive software, only the top level entity had such
a charset parameter which is good as it is a text/plain entity, but the
list archive software insists on adding it to all attachments...) you
will find that there is no Content-Encoding header, as a result the
document is not well-formed, regardless of whether you download it via
HTTP or read it from a file system or whatever. Some broken SVG viewers
are not able to recognize this fact and render the image.

More to the point, if it is delivered with Content-Encoding set and
retrieved with a Content-Encoding unaware tool, then processing will
likely fail aswell. Whether it is charset or Content-Encoding, if you do
not communicate protocol information to tools even though they depend on
such information, they can't work properly. If you tell tools of the
Content-Encoding or the charset parameter they will work properly. Yet
you do not argue that Content-Encoding is broken, so I wonder why you
insist on asking such questions?

Received on Thursday, 25 November 2004 06:29:54 UTC