Re: nhttp - Nostr over HTTP communication

pá 12. 1. 2024 v 15:45 odesílatel Ryan Barrett <public@ryanb.org> napsal:

> Interesting! Thanks for sending.
>
> I am a bit curious about the motivation though: "struggling to solve the
> high bandwidth problem." My understanding is that neither websockets nor
> Nostr's protocol itself are particularly wasteful in terms of overhead. So
> all else equal, assuming the same Nostr client and relay(s) and operations,
> would HTTP really save that much bandwidth, if any?
>

Hey Ryan.  Great to see you here.  This is a great question.

We're in new territory with nostr's architecture and scaling wss.  We need
more data, and empirical experience before we can really make good
statements about it.  Though gathering some of that data might be a useful
exercise here.

HTTP vs WSS is a really good topic.  We know HTTP scales from 30 years of
experience.  What are some of the things that make it scale?  Caching.
Content delivery networks.  Horizontal and vertical scaling through
economies of scale.  For example reverse proxies.  Stateless architecture
is widely believed to be one of the things that scale http.  In essence the
items set out in Fielding's REST, chapter 5 and elsewhere have stood the
test of time. [1]

Nostr is much more of an early experiment.  It doesnt have the same
properties, but has other ones.  It doesnt have etags, but has even ids.

So an HTTP solution can offer a tried and tested architectural style to
augment what exists on nostr, take some of the load of servers, and perhaps
scale it.  Time will tell!  I'll note that activitypub does inherit some of
the scaling benefits of http, and that appears to be doing quite well as a
network, so far.  Of course, inbox/oubpox + HTTP POST, while benefits from
clean separation of concerns, is a slightly different scaling model to a
typical REST GET system.

I think it's good to try different things, and learn from each other,
figure out where the walls are, and try to get through them.

https://ics.uci.edu/~fielding/pubs/dissertation/rest_arch_style.htm


>
> On Thu, Jan 11, 2024 at 10:32 PM Melvin Carvalho <melvincarvalho@gmail.com>
> wrote:
>
>> May be of interest to some here, though of course wss is indeed an http
>> upgrade protocol, quote from Yonle:
>>
>> "We all know that connections to #nostr relays are mostly done via
>> websocket.
>>
>> And then i got an idea: How about i create a way to communicate with
>> Nostr relays via HTTP instead?
>>
>> You see, We have QUIC already, And we know that most social media apps
>> were using HTTP to communicate. But we are using websocket. So what's wrong
>> with it? As a mobile data user, I am struggling to solve the high bandwidth
>> problem.
>>
>> So there we go, an prototype nhttp (Nostr over HTTP). It got basic stuff
>> already: REQ, and event publishing"
>>
>>
>> https://nostr.at/npub1x3azxuysp5vmfer4vgs4jn5tmfcx4ew8sh0qnev7gczljxsr7jwqa3g4el
>>
>
>
> --
> https://snarfed.org/
>