0

This is a slightly orthogonal impediment for SPDY. I have been searching for suitable articles or research conducted around this specific concern in vain.

Important prerequisite for SPDY is persistent-connection. All major sites which have very sizable wireless mobile client access switching to SPDY would mean that benefits of SPDY could potentially get lost. Here is the reason why - most of packet-based mobile networks are not capable of maintaining a resilient TCP connection mechanism across the black-spots and weak-signal areas of radio networks. If that is the case, what exactly is the point of SPDY adoption anyway ?

I have not seen yet any report on the measurable benefits of SPDY in wireless networks. Has anybody seen any ?

user1511595
  • 77
  • 1
  • 8

2 Answers2

1

https://developers.google.com/speed/articles/spdy-for-mobile

sbordet
  • 16,856
  • 1
  • 50
  • 45
  • I have seen this link before. The fact that this was a simulated network discouraged me from taking this very genuinely. However, I am planning to download Android client to test the same with an experimental server enabled with mod_spdy connecting over a 3G network. Anyway, thanks for the response – user1511595 Jul 09 '12 at 12:20
0

These are orthogonal concerns. If the physical layer is unable to deliver a payload, then the software above it won't be able to do any miracles either - that's not what SPDY is about. What you've described is, of course, a valid concern, but more so for your mobile ISP, then it is for SPDY.

If we assume that you live in a poor coverage area, and your phone and your tower are constantly dropping the signal, then re-establishing a SPDY connection is no worse off than re-establishing a TCP connection (modulo SSL handshake). As such, there really is no difference.

igrigorik
  • 9,433
  • 2
  • 29
  • 30
  • I think it might have helped if the post didn't talk about performance over mobile networks while not using a mobile network! That said I wonder if the test actually understates SPDY's advantage i.e. a single multiplexed connection vs multiple TCP connections and domain sharding. – Andy Davies Jul 09 '12 at 17:25
  • The trouble is that there is no definitive "mobile network". You can be connected to the same tower with same provider, but be on a different floor of the building and see dramatically difference performance. – igrigorik Jul 09 '12 at 19:04
  • I guess with mobile Chrome in the wild and supporting SPDY and Navigation Timing we'll at least get some real world data. Using operators test networks might be the other way to run some scenarios (if they'll co-operate of course). – Andy Davies Jul 10 '12 at 13:41