This sort of thing would be a whole lot easier if we had dmr's 
streams (not to be confused with the evil abortion that ended up in 
System V) in NetBSD. They're perfect for connection-oriented 
networking.

Sounds like you have to do FTAM over OSI CONS on top of BRI ISDN. Nasty.

If it's easier to make an X.25 implementation work in userland, then 
by all means start there. If we need to move it inside the kernel for 
efficiency, performance, code sharing, whatever, we can do that 
later, after we have an implementation that is known to actually work.

However, a word of warning from this old protocol geek - the CCITT 
(excuse me, ITU-TSS) X.25 standards documents (unlike the TCP/IP 
standards) do not specify a wire-line protocol as such; they specify 
an interface (the network, after all, is owned by PTTs who don't want 
you to know what they do behind that interface). As if that weren't 
bad enough, the language in the documents is awful - like legalese 
translated from some alien language, poorly.

To make an X.25 implementation really work, you're going to have to 
test against existing implementations for interoperability, and 
adjust your code for variances in the other implementations.

Fortunately, the last time I had anything to do with an X.25 network 
was 1996, so take some of my comments with a grain or two of salt - 
things may have improved in the interim.

	Erik <fair@clock.org>