[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: fragmenting the discussion space



I recall all the mail.  It is just my view that a shim to support SCTP
with TCP is cheaper, less cost, and less work than a shim in my kernel
as one developer.  Lets not go there I agree. I will watch this shim
discussion technically.  I don't think it will happen in the market.

/jim 

> -----Original Message-----
> From: Iljitsch van Beijnum [mailto:iljitsch@muada.com] 
> Sent: Tuesday, March 15, 2005 3:55 PM
> To: Bound, Jim
> Cc: shim6
> Subject: Re: fragmenting the discussion space
> 
> On 15-mrt-05, at 15:58, Bound, Jim wrote:
> 
> > I still believe if SCTP was deployed problem is solved.
> 
> Ok I'm going to shut up about the mobility stuff now, but I 
> have to say 
> something about SCTP...
> 
> Let's assume for a moment that SCTPs multihoming is good. (I seem to 
> remember that it has issues, but can't remember those off the 
> top of my 
> head.)
> 
> The trouble is that SCTP is as good as undeployable. First of 
> all, it's 
> incompatible with TCP, so hosts would have to know in advance whether 
> to set up an SCTP or a TCP session. Worse, the API is quite different 
> from TCP. There is some glue that makes SCTP look pretty 
> close to TCP, 
> but then you lose all the new features, so there is no point. 
> So moving 
> to SCTP means updating ALL applications, and update them in a 
> significant way.
> 
> We've been over all of this and more in multi6. The shim is the place 
> to be. I'll be happy to forward the past two years worth of messages 
> (some 3700) to anyone who wants to catch up...
> 
>