Re: [hybi] Frame size

[Ian Hickson <[e-mail]>]

On Mon, 19 Apr 2010, Thomson, Martin wrote:
> > >     this.buffer.send(prng.getNextBits(WS.FRAME_MARKER_LEN),
> > WS.FRAME_MARKER_LEN);
> >	
> > I don't understand how this solves the problem of people mismeasuring 
> > strings. (I don't really follow the pseudocode above, either.)
> 
> It does this (or significantly reduces the changes of this) by marking 
> the start of a frame with a known sequence.  This sequence could be 
> fixed, or it could be derived from a known sequence.  If this sequence 
> isn't found where it should be, there is an error.

Oh, I see. I think the behaviour of the protocol in the case of a length 
mistake today is pretty fatal already -- no need to add more catching 
code. (You get a bunch of invalid frames.) The best a client could do is 
abort anyway, and that doesn't seem especially graceful. (Ignoring the 
unknown frames is what the protocol does now.)

I thought your proposal was something that would reduce the chance of the 
error being made in the first place. That would be a good way of removing 
the need for sentinel-based framing.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'