James Carlson wrote:
Nicolas Williams writes:
For user-land applications on the send side, if the API is designed for
it, you can get away with ZC on send by having the kernel take ownership
of a user-land app's buffers on writes.
The problem is in requiring an application rewrite in order to get the
new features. Performance tweaks that require rewrites haven't had a
good history for deployment, and in the short term it runs you into
the odd questions like "so, why not do SDP instead?"
Anyway, yes, I agree that if bending applications is on the table,
then many more things become "easy."
Jim isn't this being proposed by the extended socket interface and the
proposal to use fbufs ?
Rao.
_______________________________________________
networking-discuss mailing list
[email protected]
