Awesome stuff. I love how Git just keeps getting better.
Funny (to me) aside:
"Grack [the Rack-based Git Smart HTTP process] is about half as fast as the Apache version for simple ref-listing stuff, but we’re talking 10ths of a second."
When optimizing my web app, reducing request time by 100s of milliseconds (er, I mean, "10ths of a second") would be a monumental occasion. I dance around the room when that happens. Am I wrong in thinking "10ths of a second" might be a dramatic difference for a place like GitHub?
you have to remember that this is not serving web pages - this is fetching and pushing data, which generally takes several seconds or even minutes. Having a clone take 45 seconds isn't really a big deal, but nobody would wait for a webpage load that long - it's a very different beast.
The overhead of a few hundred milliseconds is unnoticeable in almost all cases.
The user experience of pulling a git repository isn't really comparable to building a website. Also, the effect of a slow web server compounds with each additional resource on the page - as far as I can tell, the added efficiency of the new HTTP code tries to combat exactly that repeated back-and-forth.
Funny (to me) aside:
"Grack [the Rack-based Git Smart HTTP process] is about half as fast as the Apache version for simple ref-listing stuff, but we’re talking 10ths of a second."
When optimizing my web app, reducing request time by 100s of milliseconds (er, I mean, "10ths of a second") would be a monumental occasion. I dance around the room when that happens. Am I wrong in thinking "10ths of a second" might be a dramatic difference for a place like GitHub?