W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2008

Re: must a partial response range be exact?

From: A. Rothman <amichai2@amichais.net>
Date: Thu, 02 Oct 2008 10:44:42 +0000
Message-ID: <32295761.10281222944195741.JavaMail.SYSTEM@endor>
To: Julian Reschke <julian.reschke@gmx.de>
CC: ietf-http-wg@w3.org
It's a theoretical question - I've never seen this feature in use 
personally (and wasn't aware of Acrobat's use either). I'm just working 
on a small very-lightweight embeddable HTTP server, and I want it to be 
RFC compliant - so I'm raising those points in the spec which are 
ambiguous/undefined/self-conflicting, in order both to clarify how my 
server implementation should behave, and hopefully have the RFC wording 
itself modified to be clearer, thus making sure everyone who follows it 
understands the same thing (that's what specs are for :-) ).

Specifically, I will currently only support the simpler Content-Range 
responses (no multiparts), for both single-range and multiple-range 
requests, and just wanted to make sure that doing so is legal - I agree 
that it makes sense that it should be legal, but making sense is often 
not good enough ;-)



Julian Reschke wrote:

> A. Rothman wrote:
>> If a client requests partial range(s) using the Range header, is it 
>> legal for the server to return a partial result which is larger than 
>> the requested range(s) but fully contains it? This would allow a 
>> server to satisfy a request for multiple ranges by returning a 
>> Content-Range response with a single range that contains all of the 
>> requested data, thus allowing a light-weight server to be able to 
>> support partial content retrieval without having to implement the 
>> added complexity of multipart/byteranges handling and range-skipping.
> I would say it's always legal, as the response is self-descriptive.
>> It is not apparent from the spec whether this is allowed or not - 
>> perhaps it should be stated explicitly to avoid confusion (by both 
>> the sender and recipient of the response).
>> If it is not allowed, the alternative for such a server would be to 
>> support single-range requests using Content-Range partial responses, 
>> and ignore any multiple-range requests by always responding to them 
>> with full content (200) - this is implied to be legal according to 
>> the spec since range support is optional in the first place - but 
>> would be much less efficient.
>> So - must a partial response range be exact or not?
> Do you have a specific use case in mind? The only real-world scenario 
> I have seen where multiple parts were requested was with Adobe Acrobat 
> (which, as far as I recall, asks for content from the beginning and 
> from the end (the toc?))? That case doesn't seem to benefit a lot from 
> the proposal you made...
> Are there other clients you're thinking of, or is this a purely 
> theoretical question?
> BR, Julian
> ------------------------------------------------------------------------
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com 
> Version: 8.0.173 / Virus Database: 270.7.5/1702 - Release Date: 01/10/2008 09:05
Received on Thursday, 2 October 2008 11:47:10 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 11:10:47 UTC