On Fri, Apr 17, 2015 at 8:49 AM, Ryan Sleevi <sleevi@google.com> wrote:
>
> On Apr 17, 2015 8:37 AM, "Mark Watson" <watsonm@netflix.com> wrote:
> >
> > I have trouble understanding this comment. The problems we had, and
> solved, in nginx, were entirely generic and in no way
> "organization-specific". Anyone using nginx for the most basic web serving
> of large objects at any kind of scale would have the same issue. nginx is
> widely used and our solution will be available to anyone.
>
> Mark,
>
> It was not at all intended as a slight - though it seems you may have
> taken it as such. The challenges you solved _were_ unique to your
> organization - in as much as you made decisions about the design/deployment
> of OCA and had to solve your challenges within that framework. That
> includes changes to nginx and FreeBSD, but it's just as likely some other
> provider could be using Apache, or they could be serving with significantly
> beefier machines than OCA devices, but be stymied by CDN costs, or perhaps
> an inability to update devices, or any number of organization-specific
> challenges, even if the solutions are open.
>
> That you found a solution that works for Netflix's environment is great,
> and underscores the many remarks along the way that it is not TLS
> intrinsically that is the issue. Other organizations may face different
> challenges than Netflix - or, out differently, Netflix's solutions may not
> be appropriate for their environments - and having a migration path is
> exceptionally helpful.
>
Yep, other organizations will have the same, similar or different
challenges to Netflix, but essentially still stemming from the change in
workload and availability of tools / optimizations. I interpreted your
comment as saying only Netflix faces such challenges but it's great to have
recognition that such issues will be common and should be planned for. It's
not TLS specifically, but it is intrinsically the change in the nature of
the workload and the availability of technology.
...Mark