Re: [hybi] Proposed way forward for WebSockets

["Martin J. Dürst" <[e-mail]>]

As many others have said, Ian's proposal is a big step in the right 
direction.

One comment on Greg's comments below:

On 2010/07/27 13:56, Greg Wilkins wrote:
> Ian,
>
> I'm generally in agreement with the outline of your plan - specifically I
> think it
> worthwhile to get a minimalistic solution deployed sooner rather than later.
>
> But a couple of points in-line below:

> On 27 July 2010 10:56, Ian Hickson<[e-mail]>  wrote:

>> There are a number of ways that we can make the protocol also support
>> other goals:

>>   - Sending and receiving binary data

> With regards to the support for binary data, I agree that can mostly wait
> until
> the API is in javascript to handle it.  But I would also like to have one
> final discussion
> to see if we could agree to support only a single framing type that supports
> both
> binary and text - rather than enshrining multiple framing types for all
> time.
> For example,  we could switch to only supporting binary data with the
> assumption
> that unless otherwise specified the binary content is UTF-8 encoded text.

In my opinion, to put it mildly, this is a recipie for disaster. Having 
frame types for UTF-8 and binary separately creates a clear situation. 
Something like "UTF-8, but occasionally binary", or whatever in that 
direction, essentially means "any character encoding goes", which means 
that the "UTF-8" part is just meaningless. This means that we are back 
to unidentified byte gunk, where it's totally unclear what each byte 
means in terms of characters.


Regards,   Martin.


-- 
#-# Martin J. Dürst, Professor, Aoyama Gakuin University
#-# http://www.sw.it.aoyama.ac.jp   mailto:[e-mail]