couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adam Kocoloski <kocol...@apache.org>
Subject Re: Batched HTTP Requests, The I-D
Date Thu, 11 Jun 2009 12:19:54 GMT
Yes, they certainly do.

Snell's draft explicitly says this is not meant to replace pipelining,  
but there sure are a lot of similarities.  Pipelining also allows for  
multiple requests to be sent over a persistent connection without  
waiting for a response, so it achieves many of the benefits of  
Multipart/Batch.  It's interesting that Snell's example show multiple  
POST requests being batched together -- pipelining non-idempotent  
methods is something a client SHOULD NOT do per RFC 2616, Section  
8.1.2.2.

One might think of Multipart/Batch as an alternative to _bulk_docs,  
but then
> Applications processing Multipart/Batch parts MUST NOT assume that  
> any relationship or dependency exists between the individual parts  
> of the batch message.
that kinda screws with the all_or_nothing option.

Adam

On Jun 11, 2009, at 4:22 AM, Rory McGuire wrote:

> This allows submission to multiple URLs. Do HTTP/1.1 persistent
> connections allow that?
>
>
> On Wed, 2009-06-10 at 21:24 -0700, kowsik wrote:
>
>> How is this different from HTTP/1.1 which supports persistent
>> connections + multipart MIME uploads which allows for a whole bunch  
>> of
>> MIME entities to be sent in the same request?
>>
>> I didn't read the draft, but looking at the examples (and the blog
>> comment) I'm not sure what this is trying to solve and why
>> multipart/batch is required.
>>
>> K.
>>
>> On Wed, Jun 10, 2009 at 8:20 PM, Noah Slater<nslater@apache.org>  
>> wrote:
>>> Hey,
>>>
>>> This blog post from James Snell should be interesting:
>>>
>>> http://www.snellspace.com/wp/?p=991
>>>
>>> Thoughts about implementing this?
>>>
>>> Best,
>>>
>>> --
>>> Noah Slater, http://tumbolia.org/nslater
>>>


Mime
View raw message