- From: Jared Rhine <Jared_Rhine@hmc.edu>
- Date: Tue, 9 May 1995 17:04:04 -0700
- To: nazgul@utopia.com
- Cc: Multiple recipients of list <www-talk@www10.w3.org>
[Citation date: Tue, 9 May 1995 17:11:40 +0500] KH == Kee Hinckley <nazgul@utopia.com> JRhine> When I want to render an object, I first instantiate the object, JRhine> and invoke 'compute_lastmod' on it. I compare the return value JRhine> with the timestamp of an on-disk copy of the last rendering. If JRhine> they are the same, I simply serve the file off disk... KH> This has the disadvantage of leaving the user to wait until your KH> render is done. Otherwise for large options you could be presenting KH> the object while rendering, thus decreasing the overall display time KH> and increasing user satisfaction. True, but I'm not sure what your objection is. In the vast majority of cases, object rendering time is inconsequential compared to network lag and client rendering time. Being able to incrementally render client side isn't that big of a win; the time difference between streaming to the client as the object is rendering and waiting until it is done before sending is no more than a few tenths of a second at the greatest and does not substantially affect perceived performance. In any case, this model is unavoidable because in many cases, dynamic objects do not follow a linear rendering flow; that is to say, I can't properly render the top until I've processed all the data for the object. There is a possible exception for heavily loaded servers that take on the order of seconds to render objects; in this case, there would be a noticeable difference is response time. In that case, since you're working with dynamic objects, just distribute the buggers to reduce server load. Clean, simple, and cheap. Hmmm, I just considered objects that inherently take a long time to render regardless of server load (databases and such). My objects aren't like that (yet), so I haven't had to deal with it. In that case, you would indeed want some mechanism to present some feedback to the user early; in this case, we will obviously need some kind of boundary or packetized data scheme; no getting around that, I suspect, if we want accurate entity-body delimiting. -- Jared_Rhine@hmc.edu / HMC / <URL:http://www.hmc.edu/~jared/home> "I regret to say that we of the F.B.I. are powerless to act in cases of oral-genital intimacy, unless it has in some way obstructed interstate commerce." -- J Edgar Hoover
Received on Tuesday, 9 May 1995 20:06:44 UTC