Peter Memishian wrote:
> And so we have a clearview project to address snoop for loopback
> because it doesn't work as expected and...the lo0 on Solaris seems
> out of place in terms of design - an abberation, if you like.
There's nothing out-of-place or aberrant about IFF_VIRTUAL interfaces.
Except that they aren't implemented as a STREAMS device and
hence have no exposure to the rest of the system via DLPI nor
visibility in the same manner as a real network device. In a
networking stack that is meant to be based on STREAMS/DLPI,
loopback is out of place (if you like, IFF_VIRTUAL could be
considered a hack, in purist terms.) Or at least that's how they
appear to me.
> If IPMP is just more of the same then that part of Clearview is
> just perpetuating that design badness, unless we are going to use
> it as a model for others to deliver logical interfaces into Solaris?
I don't know what logical interfaces have to do with anything, nor do I
know of any "design badness" here.
Ok, ignoring the "badness" aspect (this could just be taste)...
I see nothing in the current documentation that talks about
how the IPMP interfaces fit into Solaris as a network interface.
There's plenty of discussion about how it interacts with each
command. Section 3.1 is light on detail about what form the
IPMP interface takes "behind the scenes".
Do I assume that by omission of discussion of there being a new
kernel module that the IPMP code is still integrated into IP?
> If Solaris, today, didn't have any IPMP functionality and someone
> came to you and said they wanted to build something into Solaris
> that delivers what IPMP does today, how would you recommend they
> go about it?
The same way I've proposed with the rearchitecture, without the backward
compatibility constraints.
Have you considered other approaches?
Or are there none?
I'd like to see some discussion of this in the document.
Darren
_______________________________________________
networking-discuss mailing list
[email protected]
