- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Fri, 26 Oct 2012 15:13:33 +1300
- To: ietf-http-wg@w3.org
On 26/10/2012 11:55 a.m., Adrien W. de Croy wrote: > > Hi > > ------ Original Message ------ > From: "Fall, Kevin" >> That is one case, but perhaps more interesting/controversial is this >> one: >> >> Range: bytes=100- >> >> Content-Range: bytes 200-300/1000. >> > > I'm struggling to see the point of this. > > What real world case would something like this solve? Presumably the > client would need to request 100-200 again? If not, then why even > deal in bytes and part ranges like this at all? > Squid would potentially implement this. * stale-while-revalidate and stale-if-error could use cached but not quite matching range entities to respond while the upstream request is being processed. * our "collaped forwarding" feature has some issues with late arriving requests when the start of the response entity has been discarded from buffer already. Currently we are forced to re-fetch. This would permit the some Range requests to be served despite missing chunks of entity. Amos
Received on Friday, 26 October 2012 02:14:07 UTC