At 04:26 PM 3/2/2002, you wrote: >Right. I'm content for just trying to get something to work now >and readdress this when we open 2.1. I think that there can be >multiple independent request_rec's created for one connection.
Yes - we won't revisit the idea of splitting req_rec into proto_rec and body_rec [I'm not advocating names here, just the concepts] until after 2.1. Then we can discuss how, why, etc. >I believe the concept of fast_redirect is bogus and broken. But, >you and OtherBill seem intent on keeping that. Let me be clear here... I -suggest- we keep fast redirect since it does, after all, save one entire request_rec construction/destruction cycle. Imagine; new request for /foo/ hits on mod_dir, which creates second request for index.html. Entire create_request cycle decides that it's servable. Now, if we do a 'proper' redirect, we spin that request -all-over-again- to serve it. Three request cycles, instead of two. 50% performance loss. >Based on your later messages, it seems that you are writing a >patch to add an intermediate protocol-level filters. So, I will >stay out of your way for now. Ok... we are definitely -identifying- those protocol level filters differently. The exact method we can see rbb's implementation to comment - I expect it will be reasonable. As long as insert_filter will take care of gruntwork for the module author, I'm happy.
