- From: Jason Greene <jason.greene@redhat.com>
- Date: Wed, 6 Aug 2014 09:44:54 -0500
- To: Greg Wilkins <gregw@intalio.com>
- Cc: HTTP Working Group <ietf-http-wg@w3.org>
On Aug 6, 2014, at 9:27 AM, Greg Wilkins <gregw@intalio.com> wrote: > > On 6 August 2014 23:56, Mike Bishop <Michael.Bishop@microsoft.com> wrote: > The ability to simply encapsulate the body rather than transforming it is less work for the intermediary. > > But by sending the http1 chunked body over http2 does not avoid the need to transform it. Somebody has to unchunk it and I'd really rather it not be the server. Now one might argue that those lazy browsers have heaps of available CPU, but I really don't think exporting data transform duty in either direction is a good idea. > > Besides, if it is an intermediary and not the source of the data, it has to parse the chunks anyway so it can identify the http1 end of message. It sounds like equivalent work to me. You read the chunk header, then write a data frame header. Then you use an efficient copy mechanism for the data block, repeat. If you want to reframe the chunk on different data frame boundaries you do have to buffer, but thats the same whether you encapsulate or transform the chunk stream. -- Jason T. Greene WildFly Lead / JBoss EAP Platform Architect JBoss, a division of Red Hat
Received on Wednesday, 6 August 2014 14:45:26 UTC