[You didn't cite it, but I think you're replying to an old post of mine.] Richard L. Hamilton writes: > [...] > > I'm referring to System V purists who apparently > > still think either > > that BSD sockets aren't the "real" networking > > interfaces, or that > > networking somehow is just an optional OS feature. > > Networking certainly isn't optional anymore. > > But sockets are for wimps that can't be bothered to > learn TLI/XTI. :-)
I think I'd say that TLI/XTI are too complex by at least half, and are almost completely irrelevant today -- except for being "standards" to which Solaris conforms. > (ok...so I'm kidding...maybe. ISTR a thing or two that in > principle one could do with TLI that just wasn't possible with > sockets though. I can't tell for certain, but you might be thinking of doing an early accept on TCP -- being able to "look" at the inbound connection without actually accepting it, and then reject it if you don't want it. In fact, that never did work right. The stack has always sent the SYN-ACK automatically, accepting the connection, before TLI/XTI ever gets a chance to do anything useful with it. The user level interface of TLI/XTI doesn't necessarily deliver the semantics it claims. Moreover, the lame SEQ_number design in T_CONN_IND/T_CONN_RES has complex interactions with fork and UID changes that frankly cannot be fixed. It's a shame it exists at all. > Also, the TLI stuff tends to be better documented; > I shouldn't have to buy a copy of Stevens' networking books to figure > out how to use sockets to pass an fd over an AF_UNIX socket.) I think filing a bug against the documentation would be more productive than trying to re-warm the TLI leftovers. -- James Carlson, Solaris Networking <[EMAIL PROTECTED]> Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677 _______________________________________________ opensolaris-code mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
