Re: HTTbis spec size, was: Rechartering HTTPbis

mån 2012-02-06 klockan 13:07 -0800 skrev Ted Hardie:

> "Datagram transport does not require or provide reliable or in-order
>    delivery of data.  The DTLS protocol preserves this property for
>    payload data.  "

This is a property of the underlying transport, not DTLS itself. DTLS
preserves the same guarantees (or lack thereof) as the underlying
transport.

> The major point of DTLS is that it works without a reliable transport; its
> advantages over TLS are there.

Yes.

> At least as I understand it, it also does not
> guarantee this property when run over any transport.

That's not my understanding. If the transport is reliable then DTLS is
as well.

> It might, for example,
> deliver data out of order even when the underlying transport will re-transmit
> to get in-order delivery; it depends on the implementation to determine when
> the data is passed up.

Not sure what you talk about here. in-order transports do not provide
DTLS out-of-order access to the data.

> I suspect it gets to be ignorant about the transport in part because it gets
> to assume certain things about the transport.  If you change those things,
> I think the applications that are built on top of HTTP may want new tools from
> HTTP (or they will each have to build those same tools for themselves).  This
> is particularly the case for applications where non-idempotent methods are
> used.

Well, to be honest those problems are there already even with TCP. Just
slightly different.

Please note that I do not say that HTTP/UDP is universally applicable to
all applications or even makes sense at all times. Only that UDP is one
of many transports HTTP may operate over and which makes sense for some
applications.

There is applications using HTTP/UDP today even if not part of any HTTP
standard. Not because it's cool, but because it makes sense for those
specific applications.

Regards
Henirk

Received on Monday, 6 February 2012 22:32:44 UTC