Re: HTTP over Delay-Tolerant Networks

On Fri, 31 Mar 2023, 21:49 Marc Blanchet, <marc.blanchet@viagenie.ca> wrote:

>
> Le 31 mars 2023 à 12:09, Greg Wilkins <gregw@webtide.com> a écrit :
>
> Ok. I might need help here, as I might not think on all the “special
> cases” that need to be discussed.
>


Should we have that discussion here?

It probably just needs some quick mention of existing HTTP features that
should or should not be used with DTNs.  For example:

The expect 100-continue mechanism
https://httpwg.org/specs/rfc9110.html#field.expect

This is bad for latency as it requires and extra round trip before a
request body is sent. But it is good to avoid wasting expensive bandwidth
if the server is unable to process a large request body for whatever reason.

Is section 3 the right place to explain/note such things, or perhaps the
pros/cons are self evident in the feature itself?


Http2 push https://datatracker.ietf.org/doc/html/rfc9113#name-push_promise
is a mechanism that would be very good for DTNs as it allows a server to
push associated resources without waiting for a round trip for a client to
ask for them.   However, if client doesn't want the pushed resources, then
expensive throughput could be consumed prior to the client resetting those
streams.


There are probably other features that may have specific considerations on
DTNs?


Cheers


>
>