[dtrace-discuss] [networking-discuss] Re: DTrace Network Providers, take 2

[Bryan Cantrill]

> >>When it is possible to write a dtrace script that use a probe named
> >>"ip4-physical-out-end" and there is no requirement for me to know
> >>about fbt, all will be good.
> >>
> >
> >That's only a reasonable goal if we expect customers to gather actionable
> >data from probe points of that nature. If such a probe's only consumer were
> >a developer of the IP stack, well, fbt is really the right tool.
> >
> 
> Are you saying that developers shouldn't add sdt probes and should
> only use fbt probes?

No, he's not -- Adam's point is that FBT can do quite a bit for you
without having to add SDT probes.

[ ... ]

> While we may well hide behind the veil of "sdt probes are private",
> in the great big world of open source, this is just silly.  Everyone
> can see they're there and how they can or cannot use them.  If someone
> who isn't a developer wanted to use dtrace to collect inbound packet
> statistics using dtrace and found sdt:::ip4-physical-in-start as being
> useful, who are we to say they can't if it serves their needs?

Yes, that's _exactly_ why we added a programmatic notion of stability.
So yes, you should knock yourself out adding Private SDT providers -- they
don't even require an ARC case.  And if they serve a customer's needs,
great -- because you will have declared them to be Private, a customer
that cares about the stability of their D scripts will know that their
script can be broken by any twitch of the operating system.

That said, there is great value in adding stable (or in this case, Evolving)
SDT providers -- they allow customers to build those stable scripts that
have been so useful in things like the DTraceToolkit.  

So one can envision a spectrum of usage of DTrace, with Solaris developers
on one end and end-users on the other:  FBT provides value primarily at the
developer end, Stable/Evolving SDT providers provide value primarily at
the end-user end, and Private SDT providers fall in between.  So these
are not mutually exclusive -- it's a question of the audience, and the
audience for Brendan's work is very much the end-user.

	- Bryan

--------------------------------------------------------------------------
Bryan Cantrill, Solaris Kernel Development.       http://blogs.sun.com/bmc