- From: Roy T. Fielding <fielding@gbiv.com>
- Date: Tue, 16 Oct 2007 14:58:14 -0700
- To: Mark Nottingham <mnot@mnot.net>
- Cc: URI <uri@w3.org>
On Oct 16, 2007, at 2:11 PM, Mark Nottingham wrote: > On 17/10/2007, at 3:06 AM, Roy T. Fielding wrote: > >>> Ceterum censeo: in my view the templates would benefit from an >>> easier readable syntax. >> >> Easy to read by whom? I went through the readable bits with HTTP >> and it turned out to be a big mistake. Nobody reads HTTP in real >> practice, yet the overhead of parsing HTTP messages is huge. > > I do, and my developers do; it greatly helps them understand and > debug the protocol. You don't pull out your trusty text editor and watch the stream go by -- you filter it through a protocol analyzer that maps the packets to a stream of text. The same analyzer can map the application level as well. The benefits of an ASCII protocol are a total myth aside from the framing issue: it allows most transport errors to occur unnoticed, and hence seems like it works better than a protocol that would report the error. Yes, I still use telnet to test web servers, but that is not real practice. I could just as easily test with a perl filter. Let's focus on real protocol benefits, not imaginary ones. > WRT parsing overhead -- commodity hardware can easily saturate a > gigabit Ethernet with 1K responses to GETs. "Huge" is an > overstatement. I've done the traces -- the amount of time spent looking for CRLFs in line-oriented headers is quite painful for server developers. But this has nothing to do with the topic at hand. ....Roy
Received on Tuesday, 16 October 2007 21:59:04 UTC