[ovs-dev] PORTING.md: dpif's and wildcarding

Ben Pfaff blp at ovn.org
Tue Nov 24 17:03:21 UTC 2015


On Tue, Nov 10, 2015 at 04:29:47PM +0000, Zoltan Kiss wrote:
> This documentation talks a great deal about porting strategies, and it
> mentions that "Only an ofproto provider can take full advantage of hardware
> with built-in support for wildcards"
> I wonder if it's still true, as dpif providers can use megaflows now, which
> are essentially wildcarding. Generally I think this file would worth an
> update from the core developers, as things might have changed since it's
> last modification.

It might no longer be true.

The real problem with this file is that it's all based on speculation.
No one who works with real hardware ever sends usable hardware-related
updates or patches or actionable feedback, although sometimes they send
questions ("where are the design documents for <thing>?") and complaints
("why don't you do <thing> a way that makes my life easier?").

Hardware vendors have the upper hand.  They have strictly more
information than the rest of us, because they have all of the public
information that we do and all of their private information besides.
They choose not to help.  It would be easy for them to help: they could,
for example, release specs for their hardware.  When they start helping,
hardware support and hardware information in OVS will get better.

Frankly, I have a really bad attitude about spending time trying to help
these people.  They work intentionally and deliberately to make it hard
to support their hardware.  They seem to believe that their entire
business model relies on hiding information.  Why should we, the
developers who make everything we do public, put in extra effort to
assist them?

*sigh*



More information about the dev mailing list