30 Apr
2019
30 Apr
'19
7:51 a.m.
On Apr 30, 2019, at 9:25 AM, Paul Koning via cctalk
<cctalk at classiccmp.org> wrote:
DECnet-Ultrix SOCK_STREAM sockets completely ignore message boundaries on both input and
output. Early on we ran X-windows over DECnet just by changing the socket address family
and the connect/accept logic. Very early on we had to test with SOCK_STREAM since DECnet
was the first (and only?) protocol to make use of SOCK_SEQPACKET sockets and there was a
crash lurking in the 4.2BSD kernel socket code for such sockets.
John.
On Apr 29, 2019, at 9:05 PM, Grant Taylor via
cctalk <cctalk at classiccmp.org> wrote:
On 4/29/19 6:47 PM, Zane Healy via cctalk wrote:
Yes. What I meant is that one could take an existing HTTP client and server, or create
one, substituting DECnet sockets for the TCP sockets. The protocol would work just fine
that way. You'd need to decide how to deal with DECnet packet boundaries, something
TCP doesn't have (a major omission). The simplest is to pay no attention to them,
which is what I understand Ultrix "streaming DECnet" sockets to do. An
alternative would be to make use of them, for example by saying that the entire HTTP
header is in one packet and the payload (if any) follows in separate packets.
paul
I want to say that the OSU webserver for VMS
supports running over DECnet, but my memory could be faulty. I?ve only used WASD on VMS.
I think this sounds like a neat ~> fun thing to do.
But how does a web server run over DECnet?
I guess conceptually you can serve web pages across any protocol that can carry HTTP.
But I guess you could also have a client that ran over DECnet or need a gateway to
TCP/IP.