W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2002

Re: content-encoding and range headers

From: Scott Lawrence <lawrence@world.std.com>
Date: 10 Dec 2002 09:33:51 -0500
To: Stefan Eissing <stefan.eissing@greenbytes.de>
Cc: ietf-http-wg@w3.org
Message-ID: <uznret034.fsf@world.std.com>

Stefan Eissing <stefan.eissing@greenbytes.de> writes:

> My understanding of 2616 is that range is applied to the
> content-encoded entity. So if an entity is gzipped, the
> byte selection is applied on the gzipped representation.

I don't believe that this is correct.  The Range specifies bytes in
the 'entity'; the Content-Encoding refers to a coding method
_applied_to_ the 'entity', so the Content-Range header would refer to
the bytes in the original entity, and the Content-Encoding describes
how that has been encoded to create the message body.

> What is a server to do when a client sends a request with
> Accept-Encoding: gzip
> Range: bytes=0-499
>
> Is the server free to chose the content-encoding to its liking
> (identity or gzip)?

Yes, but the result would be the first 500 bytes of the entity, sent
either with or without gzip encoding at the servers option.

> Assuming there is a second request later
> Accept-Encoding: gzip
> Range: bytes=500-1000
>
> Is the server allowed to use another content-encoding in the response?

The 'identity' encoding is the only one that is implicitly allowed, so
it shouldn't send any others.
 
> In that case, a server storing gzipped entities has some work to do
> answering range request. It would then need to "uncompress"
> (ungzip?) the entity before it can select the correct range.

This gets into an interesting question, and one that often causes
confusion.  Let's say that I have a file 'foo.txt.gz' on my web server
at URL '/foo.txt.gz'.  The (gzipped) file is 350 bytes long, but
expands to 1000 bytes when ungzipped.

 GET /foo.txt.gz HTTP/1.1
 Host: scott.example.com
 Accept-Encoding: gzip

I'll return 

 HTTP/1.1 200 Ok 
 Content-Length: 350
 Content-Type: application/x-gzip

 [350 bytes of the file]
 
Note that I did not send a Content-Encoding header, because the entity
itself is gzipped, which is expressed in the Content-Type.  

If the file were stored as 'foo.txt' and you sent:

 GET /foo.txt HTTP/1.1
 Host: scott.example.com
 Accept-Encoding: gzip

then I might compress the response and send:

 HTTP/1.1 200 Ok 
 Content-Length: 350
 Content-Type: text/plain
 Content-Encoding: gzip

 [350 bytes]

--
  Scott Lawrence
Received on Tuesday, 10 December 2002 09:33:58 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:49:21 GMT