[Info-vax] OpenVMS app stacking

[Stephen Hoffman]

On 2018-01-03 20:03:04 +0000, DaveFroble said:

> However, one part reminds me of a discussion we had in the past.  You 
> didn't seem too thrilled with my idea of a listener which would peek 
> into a shared socket buffer, then send (in whatever way appropriate) 
> the work to another app which would us the same socket.  Think that 
> idea might work for all those apps waiting for HTTPS connections?  I'm 
> guessing it might require some work on the destination apps.

The implementation of an application-level TLS router is something I'd 
prefer to avoid.

The Auxiliary Server (what most folks call inetd) and/or borrowing the 
code in Apache port.  Or see how portable something like Gearman 
http://gearman.org might be.

Pushing the balancing toward the client where that's feasible, or to 
load balancing gear where that's available.   Tell the client to 
directly chat with {other end-point} for the next {time}.  Or program 
the client to use DNS SRV records, and load-balance those with the TTL 
de jour.

DTLS where connections aren't needed or 2PC retries can be implemented locally.

Maybe DTLS with MQTT or CoAP, or move to TLS with a REST library or 
with Tomcat or such.

For designs where there's enough system performance available, open a 
chain of TLS connections.

I'm not a huge fan of trying to pass around the entire context of a TLS 
end-point.   That hand-off seems unlikely to be reliable.  Or something 
akin to MiTMproxy https://mitmproxy.org and which I'd prefer to avoid 
in production designs.



-- 
Pure Personal Opinion | HoffmanLabs LLC