Remy Maucherat wrote:
Look, there's no reason to start whining. Adding a lower level buffering (using a ByteChunk maybe) is nearly trivial, and since you seem to know what you're doing, you could have done in less than the amount it took you to write your email. I'm not convinced it's the best solution all the time (Coyote leverages the buffering mandated by the higher level - the servlet API - to avoid useless copying from buffer to buffer).
-1 for option 2 (I consider keepalive is more important than anything
else).
I am perfectly willing to work on this, but it did not seem trivial to someone not familiar with the coyote connector. I need to be able to point out issues (and find out if they really are problems) without being accused of whining.
Well, I thought the code which outputs the bytes would be simple enough. I will experiment.
I'm not thrilled on buffering (the compression stuff should probably be the one doing more buffering, and I can add a simple trick - in most situations - to the chunking algorithm to prepend the chunk header to the chunk data buffer).
You seem to imply that keepalive is tied to chunking - I believe these are independent and chunking could be disabled without sacrificing keepalive.
This is not correct. Most JSPs can't keepalive without chunking. And many servlets are the same (a lot of people think doing out.flush followed by out.close when finishing handling the request is a good idea).
Remy
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
