My $0.02 on SPDY

November 15th, 2009 Leave a comment Go to comments

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

The whole online community is a-buzz on the experimental SPDY protocol (pronounced Speedy) recently proposed by Google.  I’ve been holding myself in since I heard about it, and want to set the record straight on a few things.  I know that a lot of people from my marketing blog won’t really get this, and a lot of traffic stumbling on this won’t appreciate the rest of my blog (which just kinda sucks for you guys) but here’s my geeky response to all of the hubbub.

First of all, please stop writing about how SPDY is going to replace HTTP.  It’s not.  It’s mostly  going to wrap HTTP.  RTFM folks, it says so black and white in the 2nd question in the FAQ there.

Second of all, the server push, and server hints have been something that I’ve been talking about for years.  I’ve had some conversations with Roy Fielding about that (go Google his name if you don’t know it) and the answer was always to not do it.  There are very few cases where it will really help and in most of those cases that will require caching and/or pre-processing of the original or primary resource server-side.

Third of all, the Apache HTTP Server community addressed the lengthy headers a years ago.  A replacement protocol called WAKA was mentioned.  I’ve yet to see someone pick up that gauntlet but if the folks at Google care so much, why don’t you use that instead of adding mandatory compression around the whole protocol.

Fourthly, to all you small webmasters who are drooling over the speed boosts, did you notice that every web site now needs SSL?  The IP isn’t a problem, as I assume that SPDY-enabled servers and clients will also be SNI-enabled, but that does mean shelling out money for signed certs on every single website, based on my understanding of the SPDY proposal.  And if the server/clients are NOT forced to be SNI, then SPDY is going to pollute the IP space – which might actually be a good thing if it forces more widespread adoption of IPv6 ;)

Finally, I have to wonder about the timing of the announcement.  The fact that there are already beta clients and servers suggest that announcing this a few weeks earlier would likely have been possible.  If so, why deliberately wait until a week AFTER the larget opensource HTTP server’s semi-annual conference to announce it, instead of flying Google’s top folks over to Oakland to talk about it with some of the biggest people in the webserver community?  It just seems… odd… to me.

In conclusion, I think that SPDY has potential.  Google can “pull a Microsoft” and shove the protocol down our throats one way or another if they really want to, but is it really justified?

  1. No comments yet.
  1. No trackbacks yet.