- From: Brian Behlendorf <brian@wired.com>
- Date: Wed, 12 Oct 1994 12:17:25 -0700 (PDT)
- To: Henrik Frystyk Nielsen <frystyk@w3.org>
- Cc: http-wg@cuckoo.hpl.hp.com
On Wed, 12 Oct 1994, Henrik Frystyk wrote: > A general problem that I see in the proposals until now is that we have > no guarantee that the images are in fact on the same server as the main > document. Often this is _not_ the case and then it doesn't help to keep > the connection open nor is it easy for the server to get the size of > the image. Actually, I'd dispute this. I bet we could get one of our web-crawler authors to add to his crawling algorithm a measure of the ratio of inlined-images-on-same-site to inlined-images-off-site, and that it would probably be something on the order of 20-1, if not 100-1. We can't guarantee it, and we certainly shouldn't set up a protocol that would make inlining off-site images difficult or impossible, but forgoing optimizations because of it is a bad choice, I think. > I think a general solution must be based on at least two connections. > First the main document gets retrived. If the client is text-based then > fine - no more connections are made. If not then the client can sort > the requests for inline images and make simultaneously (multi-threaded) > connenctions to the servers involved. These can then be multipart, MGET > or whatever solution we come up with. Right, this would be great, and I don't see how it contradicts other proposals made here, it's just another parallel action. Brian
Received on Wednesday, 12 October 1994 20:17:24 UTC