[Top] [All Lists]

Re: [ietf-smtp] [Proposal] confusing parts of the mail system, was 250-MARKDOWN

2019-01-16 15:16:35
In article <0826bb60-8ebb-f531-2a1d-8fad91335ff1(_at_)wizmail(_dot_)org> you 
On 16/01/2019 20:02, Gene Hightower wrote:
On 16/01/2019 11.15, Jeremy Harris wrote:
Efficiency.  Not having to do dot-stuffing means that sendfile etc.

Can you use sendfile() after STARTTLS?

I'm not aware of the common TLS libraries providing an equivalent
interface yet, but once the symmetric-crypto portion is established
in the kernel I'd think it possible.  Perhaps five years from now?

Even if someone puts some of the crypto glop in the kernel, the
benefit of sendfile() is minimal.

The point of sendfile is that the kernel can send network data
directly from the disk buffer.  That's never going to happen if it has
to be encrypted first.  If it has to read the buffer, encrypt it, and
put it back, that's just like copying it while encrypting.

Either way, it seems like a pretty minor optimization.  My MTA
routinely has a hundred connections going at once and it's totally I/O


ietf-smtp mailing list

<Prev in Thread] Current Thread [Next in Thread>