One thing to remember is Base64 encoding is less efficient than the raw binary. Also, all modern web servers and browsers support reusing TCP connections for multiple HTTP requests (though there is still the extra latency for each actual request, just not the TCP connection sequence, etc)
I'd be interested to see actual benchmarks, rather than this "trust me, it works" thing.
The Base64 issue should be relatively minor as long as you're gzip-encoding everything your server sends. (the iPhone supports that, right?) Each base64 byte only has 6 bits of entropy so it'll compress well. I'd be more worried about maintaining efficient caching with this technique.
First point: MUCH thanks to the author for this tip. The working examples definitely showed the difference on my iPhone on edge.
Second point: who in the hell cares how the word "iPhone" is capitalized? Really, is the most pertinent comment you can contribute about being a f*cking grammar Nazi?
On a fun note, the iPhone makes sure you cap the "P" automatically. It doesn't care about the caps on microsoft.
Can we please, please start referring to the iPhone as it's actually called... _i_Phone, not Iphone or IPhone or I-Phone. Lowercase i, uppercase Phone.
iPhone.
I'm starting to think that News.YC does this automagically... someone help a brother out here.
Specifically: The script assumes that words with capitalized letters other than the first character are already correctly capitalized. This means it will leave a word like “iTunes” alone, rather than mangling it into “ITunes” or, worse, “Itunes”.
I'd be interested to see actual benchmarks, rather than this "trust me, it works" thing.