Re: [hybi] hum #3: Message

[Scott Ferguson <[e-mail]>]

Ian Hickson wrote:
> On Thu, 5 Aug 2010, Salvatore Loreto wrote:
>   
>>    * A message may be composed of one or more "frame"s
>>    * Each frame has a length indication, encoded in a fixed number of
>>      bits (where that lengths is fixed in the specification to be written)
>>     
>
> When transfering large blocks of data, chunking is quite inefficient. This 
> means that we'd need to make sure that the fixed width field is at least 
> long enough to handle large chunks of data.
>   

Instead of guessing, I measured chunking vs non-chunking for large 
(16meg) HTTP requests (10000 requests measured by ab on localhost, 
single client) using Resin.

The biggest difference I saw was 3% after several runs, most were within 
1%, and in some cases the chunked beat length (yay GC.)

That was HTTP chunking (8 bytes with 4 encoded as ascii hex), a 8k 
chunk/buffer size, and zero application time (write a pre-filled 64k 
buffer 256 times). In other words, a bigger overhead than the proposed 
frame header.

In other words, the overhead of chunking is so small that it's barely 
measurable, even in the most extreme case.

-- Scott
> At terabit Ethernet speeds, 32 bits is not enough since you could transfer 
> that much in mere seconds, so presumably we'll use 64 bits.
>
> Do we really want a length that long? 
>
> The alternative is variable width lengths, which are no more complicated 
> to implement than fragmentation, and which do not preclude using 
> fragmentation either, and which remain compact at any length.
>
>
>   
>>    * A message may be composed of one or more "frame"s
>>     
>
> I think we should obtain real-world experience with how people use 
> WebSockets before we add fragmentation support to the technology. Since it 
> can easily be worked around at the application layer (by having 
> fragmentation in the subprotocol), we can easily collect information on 
> how people are using this after we have initial deployments.
>
>
> Other relevant e-mails on the subject from other people:
>
> http://www.ietf.org/mail-archive/web/hybi/current/msg02841.html
> http://www.ietf.org/mail-archive/web/hybi/current/msg02830.html
> http://www.ietf.org/mail-archive/web/hybi/current/msg02892.html
>
>