getcontentlength property

Werner Donné wrote:
> Hi,
> 
> RFC 2518 says that the getcontentlength property is
mandatory if
> the resource returns the Content-Length header in
response to a GET.
> 
> In a servlet container you don't always know if that
header will
> be in the response. Normally a document is streamed to
the response
> body. The container sees this because of the
intermediate flushes.
> It will then use the chunked transfer encoding. It can
be, however,
> that the document is smaller than the internal
streaming buffer.
> There are no intermediate flushes then and the
container may choose
> to not use a transfer encoding, in which case the
Content-Length
> header will be set.
> 
> It is also an expensive operation if the document is
versioned and
> must be reconstructed from a delta file.

Agreed, there are cases where a WebDAV server implementor
simply doesn't 
have any control over this. Therefore the MUST level
requirement should 
be relaxed to a SHOULD.

Best regards, Julian

(cc:ing the WebDAV WG mailing list, because this happens not
to be a 
Versioning question)

Julian Reschke wrote:
> 
> Werner Donné wrote:
>> Hi,
>>
>> RFC 2518 says that the getcontentlength property is
mandatory if
>> the resource returns the Content-Length header in
response to a GET.
>>
>> In a servlet container you don't always know if
that header will
>> be in the response. Normally a document is streamed
to the response
>> body. The container sees this because of the
intermediate flushes.
>> It will then use the chunked transfer encoding. It
can be, however,
>> that the document is smaller than the internal
streaming buffer.
>> There are no intermediate flushes then and the
container may choose
>> to not use a transfer encoding, in which case the
Content-Length
>> header will be set.
>>
>> It is also an expensive operation if the document
is versioned and
>> must be reconstructed from a delta file.
> 
> Agreed, there are cases where a WebDAV server
implementor simply doesn't 
> have any control over this. Therefore the MUST level
requirement should 
> be relaxed to a SHOULD.
> 
> Best regards, Julian
> 
> (cc:ing the WebDAV WG mailing list, because this
happens not to be a 
> Versioning question)

Added as issue to BugZilla: 
<http://ietf.cse.ucsc.edu:8080/bugzilla/show_bug.c
gi?id=239>.

Best regards, Julian