Re: HTTbis spec size, was: Rechartering HTTPbis

2012/2/6 Henrik Nordström <henrik@henriknordstrom.net>:
> mån 2012-02-06 klockan 12:05 -0800 skrev Ted Hardie:
>
>> Would you propose different method names for this mode, so that they
>> were distinguishable
>> to intermediaries, or would propose that the semantics of the method
>> be determined by
>> the transport on which it arrived?
>
> I would simply state that there is no guaranteed delivery status when
> HTTP/UDP is used. Messages MAY be silently lost when using such
> transport and it's up to to the application to handle it gracefully.

There's also no guarantee that all messages will fit into a single UDP packet,
so you may get partial messages that need to be discarded.

>How
> is outside of the specification. There is also other very noticeable
> limitations of HTTP/UDP such as maximum message size which is fairly
> small. I.e. 64KB on a good day.
>
> If reliable datagram transport is desired then use HTTP/SCTP.  It's a
> very real alternative in the environments and for the applications where
> HTTP/UPD is interesting and provide much the same properties in
> unordered (but well defined) delivery of interleaved messages and other
> nice properties.
>
Note that many of the deployments of SCTP currently run over UDP.

regards,

Ted



> Regards
> Henrik
>

Received on Monday, 6 February 2012 20:57:41 UTC