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

Re: Last gasp terminology issue

From: <hallam@etna.ai.mit.edu>
Date: Wed, 05 Jun 96 12:18:10 -0400
Message-Id: <9606051618.AA03230@Etna.ai.mit.edu>
To: Daniel DuBois <dan@spyglass.com>, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Cc: hallam@etna.ai.mit.edu

At 09:50 AM 6/4/96 -0700, Roy T. Fielding wrote:
>> 14.24 If-Modified-Since
>>  
>> The If-Modified-Since request-header field is used with the GET method
>> to make it conditional: if the requested variant has not been modified
>> since the time specified in this field, an entity will not be returned  ...
>
>The above is a protocol error.  IMS applies to the resource as a whole,
>not to any single variant.  It should be

I fail to see the cost/benefit analysis of this move.

In the first place I know of very few servers with a genuine understanding of 
how to process variants. In CL-HTTP there is genuine format translation on the 
fly and caching of previously calculated translations. In this case I would 
consider it obvious to use the date at which the original resource was generated 
rather the date on which the variant was translated.

If on the other hand we have a CERN server type setup where one is simply 
serving out the file system and translation on the fly is a post facto hack I 
don't see that the content negotiation argument holds. I don't beleive that 
anyone is going to search through an entire directory to discover the original 
version. Indeed there is no way for a CERN/Apache type server to know which was 
the origin format. Guessing that the oldest variant is the origin sounds 
dangerous to me.

I think that this proposal attempts to go beyond the technical capability of 
almost every server with a significant user base. Unless one is backing the 
entire filestore as Jigsaw/CL-HTTP and like servers do it will not be possible 
to perform the operation efficiently. Moving to a backing store architecture is 
a major proposition and many operating systems simply could not support high 
load with that architecture, it requires a single process - multithreaded 
server.

Dan is entirely right on this one - NO.

		Phill
Received on Wednesday, 5 June 1996 09:16:18 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:02 EDT