Hi Jim, Mario

I do not dispute that stateful interactions over HTTP are possible.
Mechanisms such as cookies (RFC6265) clearly allow applications layered on HTTP 
to maintain state.

However, the existence of such mechanisms does not necessarily make them an 
appropriate fit for every protocol. The question here is not whether HTTP can 
support state, but whether introducing HTTP-layer session semantics into EPP 
leads to a clean and well-defined protocol architecture.

In RFC 5730 there is a clear separation between:

the EPP protocol itself, which is stateful and defines its own session 
semantics (login, logout, command sequencing), and the transport layer, which 
RFC 5734 currently specifies as TCP with TLS.

If EPP were mapped directly onto HTTP while preserving EPP’s existing session 
model, the result would effectively introduce two overlapping session layers:

- EPP session state (login/logout, command sequencing)
- HTTP session state (cookies or similar mechanisms)

At that point the transport is no longer a simple framing mechanism, but begins 
to influence protocol semantics. That tends to complicate both implementations 
and interoperability, because state transitions must now be coordinated across 
two layers.

With EPP over TCP, the behavior is straightforward. If the client stops sending 
requests, the TCP connection closes, and the EPP session dies immediately. EPP 
over HTTP is messier. The session now depends on the lifetime of the HTTP 
connection or session, which means the server must keep EPP session state alive 
somewhere. This can result in stale or abandoned session data lingering for a 
long time, increasing complexity and resource usage, and making a clean, 
predictable session model much harder to maintain.

There is also a practical scope concern for the working group. If HTTP is 
introduced as a transport, even in a minimal form, there will inevitably be 
pressure from implementers to make use of additional HTTP features. for example 
cookies, caching semantics, additonal headers, additional authentication 
schemes, or other aspects of the HTTP ecosystem.

Each of these raises new questions about how they interact with EPP semantics 
and whether they should be specified, restricted, or ignored.

This can easily lead to the working group spending significant effort defining 
the boundaries of HTTP usage rather than focusing on protocol extensions. 

Production deployments demonstrating that EPP over HTTP can work are valuable 
input. However, they do not necessarily resolve the design question of whether 
the resulting architecture is the clearest or most robust model to standardize.

I would be cautious about using DNS as an example for "modernizing" EPP. DNS is 
fundamentally a stateless protocol and therefore maps much more naturally onto 
HTTP-based transports.

My main concern is therefore not whether EPP over HTTP can function in 
practice, but whether standardizing it in this form results in a transport 
mapping that remains consistent with the layering model established in RFC 5730.
_______________________________________________
regext mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to