W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1996

Re: 3.17.2 Byte Ranges

From: Jeffrey Mogul <mogul@pa.dec.com>
Date: Thu, 25 Apr 96 14:25:28 MDT
Message-Id: <9604252125.AA15650@acetes.pa.dec.com>
To: hardie@nasa.gov
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
X-Mailing-List: <http-wg@cuckoo.hpl.hp.com> archive/latest/342
    In Section 3.17.2, the spec says that the if a last-byte-pos value is
    larger than the current length of the entity, it is assumed to be
    equal to the current length of the entity.  From my reading, a server
    would thus send first-byte-pos to the end of entity.
I think you are referring to a case like this: /foo.gif is 1024 bytes
long, and a client sends
	GET /foo.gif HTTP/1.1
	Range: bytes = 0-2047
The spec now implies that the returned response would have
	Content-Range: bytes 0-1023/1024
Is this consistent with your reading?

    I'm afraid this behavior doesn't make sense to me.  It seems to me
    that this should be treated as an invalid byte-range-spec, and that
    the recipient should ignore it and send back the entire entity.
    The sender already has a method of requesting from first-byte-pos to
    the end of the entity (by leaving off the the last-byte-pos), and
    I don't understand why we want to duplicate this behavior in response
    to what looks like an error.  

I copied this part of the spec from draft-ietf-http-range-retrieval-00.txt,
which had met with general approval 9at least in that section).  If
we change it, we'll need to make sure that nobody objects.

My own feeling is that a simple application of the robustness principle
here would say "return the most helpful response", not "return an error".
After all, the response is clearly marked with a Content-Length
that allows the recipient to decide if the right thing happened.

Consider this case: the client currently has a cached copy of a GIF
file, and wants to see if the copy at the server has been updated.
However, it's a huge interlaced GIF file, so the client only wants
to see the first "layer" (or whatever it's called) initially.

So the client might send
	GET /huge.gif HTTP/1.1
	If-Invalid: "cache-validator-value-1"
	Range: bytes = 0-4095

Meanwhile, the copy at the server HAS been updated, but the service
author has figured out that the image ought to be a lot simpler, and
it now only has 2048 bytes.  Since the new validator doesn't match
the one in the If-invalid, the server would normally return a 200
(OK) response to this conditional GET, not a 304 (Not modified)
response.  In this case, it seems perfectly reasonable to return
the entire smaller file, rather than telling the poor client
207 (Range Out Of Bounds) meaning "sorry, you will have to ask me again,
even though I could have sent you something useful."

I.e., I don't think this is an error condition.

Received on Thursday, 25 April 1996 14:41:41 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:40:16 UTC