Sebastien Roy writes: > On Fri, 2008-05-30 at 17:05 -0400, James Carlson wrote: > > IS-IS uses ISO datagrams, and thus requires raw network access. > > Because we're delivering the source upstream to quagga.net, which > > compiles on many older Sun platforms, we use standard DLPIv2 rather > > than the newer OpenSolaris libdlpi convenience library. > > How does the daemon decide whether to open DLPI devices out of /dev/net > on some versions of Solaris, but out of /dev on others?
It speaks traditional Solaris DLPI, which means that it opens nodes in /dev, and knows nothing of /dev/net. What I think you're actually saying is that we (like all DLPI applications) have got a problem because of Clearview link renaming in Nevada. If that's what you're trying to say, then I agree, and I think we ought to add /dev/net/ support. As for versions, they just don't matter. If you can open a node /dev/net/, then it must already exist, which means that the platform you're running on (whatever it is) supports /dev/net/. There's no reason I can see to make this conditional on version. It should be trivial. Nicolas Williams writes: > On Fri, May 30, 2008 at 05:37:14PM -0400, Sebastien Roy wrote: > > How does the daemon decide whether to open DLPI devices out of /dev/net > > on some versions of Solaris, but out of /dev on others? > > And wouldn't it be easier to contribute libdlpi to quagga.net rather > than avoid its use? I don't think it would. We'd end up with multiple versions of that library delivered (one through quagga.net and the other through ON), and all the problems that causes. We'd be hard-pressed to keep the forked sources in sync, and make sure the right thing happens on each platform. All of the DLPI code in Quagga, including the special code for filtering out ISO packets and glue into the Quagga data structures, amounts to 600 lines of code and comments. And other than (perhaps) the debatable path issue, it's all based on stable interfaces. If we were to go this direction, I'd rather make Quagga dependent on an ON patch that delivers libdlpi to S7, S8, S9. (Not sure if S7 is old enough ...) -- James Carlson, Solaris Networking <james.d.carlson at sun.com> Sun Microsystems / 35 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677
